+ All Categories
Home > Documents > Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a...

Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a...

Date post: 28-Jul-2020
Category:
Upload: others
View: 6 times
Download: 0 times
Share this document with a friend
48
Chapter 5 Syntax-semantics interface of wh-questions In the previous chapter, we concentrated on the syntactic aspects of wh- questions. We showed how to account for cross-linguistic differences in terms of the wh-type schema proposed in chapter 3. So far, we have used wh as a temporary abbreviation for categorizing wh-questions. In this chapter, we propose to spell out the wh type by incorporating the types of possible answers into the type assigned to wh-questions. Based on the Curry-Howard interpretation of the decomposed type, we provide the se- mantic term decomposition of the semantic operator ω. With the decomposed type and term, we can determine the lexical semantics of each wh-phrase. In- terestingly, the syntactic and semantic decomposition of wh-phrases leads to derivability patterns between instances of wh-type schema. Syntactically, the derivability patterns between wh-type schemata make it possible to derive different types of wh-questions from a single wh-type schema. Semantically, the semantic term assigned to a wh-type schema results in a uniform meaning assembly of wh-questions. Using these patterns, we can account for multiple wh-questions and different types of answers from a single instantiation of the wh-type schema. Additionally, we provide syntactic analyses of different types of wh-questions, e.g. multiple wh-questions, embedded wh-questions, scope-marking constructions. In section 5.1, we discuss two approaches to the semantics of questions: the proposition set approach and the structured meaning approach. We will argue in favor of a structured meaning approach. In section 5.2, after providing the type definition of wh-questions, we show how the decomposition of the macro type wh for wh-questions leads to derivability schemata between instances of wh-type schema. Finally, in section 5.3, we explore the polymorphic use of the
Transcript
Page 1: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

Chapter 5

Syntax-semantics interface ofwh-questions

In the previous chapter we concentrated on the syntactic aspects of wh-questions We showed how to account for cross-linguistic differences in termsof the wh-type schema proposed in chapter 3 So far we have used wh as atemporary abbreviation for categorizing wh-questions

In this chapter we propose to spell out the wh type by incorporating thetypes of possible answers into the type assigned to wh-questions Based onthe Curry-Howard interpretation of the decomposed type we provide the se-mantic term decomposition of the semantic operator ω With the decomposedtype and term we can determine the lexical semantics of each wh-phrase In-terestingly the syntactic and semantic decomposition of wh-phrases leads toderivability patterns between instances of wh-type schema Syntactically thederivability patterns between wh-type schemata make it possible to derivedifferent types of wh-questions from a single wh-type schema Semanticallythe semantic term assigned to a wh-type schema results in a uniform meaningassembly of wh-questions Using these patterns we can account for multiplewh-questions and different types of answers from a single instantiation ofthe wh-type schema Additionally we provide syntactic analyses of differenttypes of wh-questions eg multiple wh-questions embedded wh-questionsscope-marking constructions

In section 51 we discuss two approaches to the semantics of questions theproposition set approach and the structured meaning approach We will arguein favor of a structured meaning approach In section 52 after providing thetype definition of wh-questions we show how the decomposition of the macrotype wh for wh-questions leads to derivability schemata between instances ofwh-type schema Finally in section 53 we explore the polymorphic use of the

154 5 Syntax-semantics interface of wh-questions

wh-type schema by reanalyzing some of the cross-linguistic data from chapter4 and we show that we can account for the correct syntactic type and meaningassembly of wh-questions

51 Semantics of questions and answers

The type-logical grammar approach imposes a strong correlation between thesyntactic construction of an expression and its semantic interpretation Thiscorrelation is formulated in the Curry-Howard correspondence between syn-tax and semantics The correspondence stems from Fregersquos principle of com-positionality The meaning of the words (= lexical semantics) and the waythey are put together (= derivational semantics) determines the meaning ofthe whole

Many theories that account for the semantics of questions relate the mean-ing of a question to its possible answers (for an overview of those theoriessee Groenendijk and Stokhof (1997)) We present two approaches of relatingquestions and answers the proposition set approach (Hamblin 1958 Karttunen1977 Groenendijk and Stokhof 1984) in which questions represent propos-itions and the approach which Krifka (2001) named the structured meaningapproach also referred to as the functional or categorial approach (Groenendijkand Stokhof 1984) In this latter approach the interrogative in combinationwith its answer forms a statement

511 Proposition set approachThe proposition set approach was formulated by Hamblin (1958) who wrote aset of principles that largely influenced the logical approach to the semanticsof questions (Karttunen 1977 Groenendijk and Stokhof 1984) With theseprinciples Hamblin (1958) captures the main idea behind a logical approachto interrogatives which states that to determine the meaning of an interrogat-ive one has to inspect what kind of statement can serve as a response The firstof Hamblinrsquos principles was paraphrased by Groenendijk and Stokhof (1997)as ldquoan answer to a question is a sentence or statementrdquo The principle imple-ments the idea that the semantic status of an answer is a proposition and thatthe syntactic form of an answer is irrelevant

Answers as statements Constituent questions such as lsquoWhat did Mary readrsquoor lsquoWho read The Minimalist Programrsquo receive a full statement as an answer Alikely answer to either question could be lsquoMary read The Minimalist ProgramrsquoAnswers which only give a single constituent as an answer (lsquoThe Minimalistprogramrsquo or lsquoMaryrsquo) are said to be derived from the full statement Thus theunderlying structure of an answer has to be a complete statement with therequested argument embedded in it

51 Semantics of questions and answers 155

The semantics of questions in the proposition set approach are representedin variants of predicate logic Adopting the presentation style of Krifka (2001)we illustrate the semantics of a question lsquoWho read The Minimalist Programrsquo inpredicate logic

Question lsquoWho read the Minimalist Programrsquop | existx(person x)and p = ((read mp) x)

= ((read mp) x)|(person x))= ((read mp) m ((read mp) j

Answer lsquoJohn read the Minimalist programrsquo= ((read mp) j

The answer to such a question is represented by the set of propositions p thatare restricted by the predicates that have to hold for this proposition For ex-ample there is a person for which it is true that this person read the MinimalistProgram By analyzing the restrictions as part of the proposition the answeryields all interpretations that make such a proposition true

The main critique of the proposition set approach is that it lacks express-ibility Because Hamblin (1958) chooses to ignore the syntactic form of theanswer and treats all answers as full statements the proposition set approachcannot directly relate a certain type of answer to a certain type of questionKrifka (2001) notes that questions with alternative answers or multiple con-stituent questions cannot be handled easily in the proposition set approachThe dominant view is that questions belong to a single category and shouldtherefore be mapped to one type This stands in contrast to the view whichis reflected by the structured meaning approach which takes a polymorphicstance towards the categorial nature of questions

512 Structured meaning approachThe structured meaning approach is sometimes referred to as the functional orcategorial approach The approach is developed by logicians and semanticistsand supports the idea that the meaning of a question is dependent on themeaning of the answer and vice versa Along similar lines Hiz (1978) pointsout that questions and their answers are not autonomous sentences but thatthey form a semantic unit mdash a question-answer pair We briefly discuss thestructured meaning approach and its syntactic consequences

Constituent answers An appropriate answer to a single constituent ques-tion may be any type of syntactic object This might be a generalized quantifierphrase or a verb phrase as well as a noun phrase or prepositional phrase Ad-ditionally in multiple wh-questions different combinations of syntactic ob-jects can be used as an answer The wh-question directs the kind of answersthat can be expected

(51) a lsquoWho saw Maryrsquo John nobody Johnrsquos sister

b lsquoWhen did John see Maryrsquo In the afternoon

156 5 Syntax-semantics interface of wh-questions

c lsquoWhich man did John seersquo His father the neighbor

d lsquoWhy did John see Maryrsquo Because

e lsquoWho saw whomrsquopair list reading John (saw) Bill Mary (saw) Sue functional reading every professorhis student Johnhis sister

As the sentences illustrate the answers have a direct relation to the interrog-ative phrase in the question To capture the relation between the question andits possible answer type the structured meaning approach formulates the ideathat the question and answer form a unit both syntactically and semanticallySyntactically the interrogative in combination with its answer forms an indic-ative sentence or a question-answer sequence This syntactic unit is reflectedin the semantics where the question meaning is a function that yields a pro-position when applied to the meaning of answer (Krifka 2001)

In the proposition set approach answers belong to a uniform type of syn-tactic object viz s whereas the answers in the structured meaning approachmay fall into any kind of category A syntactic analysis must formulate a wayto account for the possible variation in question-answer pairs Along similarlines Groenendijk and Stokhof (1984) have argued that a proper theory of thesemantics of questions should not focus on one type for wh-questions butdefine a multitude of types and determine the semantic relationship betweenthem Groenendijk and Stokhof (1997) refer to this view as ldquothe polymorphicstancerdquo

Within the type-logical grammar framework a polymorphic view on ques-tion and answer types comes quite naturally as shown in work of Hausser(1983) and more recently in Bernardi and Moot (2003) These works showthat the diversity in answer types can also be derived from uniformily typedwh-phrases We will follow this line of ldquoderivational polymorphismrdquo andshow that by incorporating the answer type into the type assigned to wh-phrases we can account for different question-answer sequences on the basisof a single type for wh-questions

52 Question and answer types

In chapter 3 we assigned type wh to wh-questions as a temporary abbrevi-ation In a structured meaning approach questions are expected to be func-tions that when applied to an answer yield a proposition In this section wewill spell out the wh abbreviation in such a way that it reflects the functor-argument relation between a wh-question and its response In section 521we determine how this relation is captured in the syntactic and semantic typedefinition of wh-questions The next step is to determine the lexical semanticsof wh-phrases In section 522 we work out the lambda term for the ω oper-ator that we have used so far as a semantic operator for wh-questions In sec-tion 523 we will show that spelling out the wh-question type along with the

52 Question and answer types 157

meaning assembly for the ω-operator leads to derivability patterns betweeninstances of the wh-type schema These patterns follow from characteristictype-shifting laws that hold in the semantic type language

521 Type definition of wh-questionsSo far we have assumed wh-questions to be expressions of type wh Adopt-ing a structured meaning approach of questions we will incorporate the typeof possible answers into the type of the wh-question Generalizing over thepossible types of answers and questions we decompose the type abbreviationwh of wh-questions into the following type

Definition 25 Decomposition of wh-question type wh

syntactic type semantic type

B A = A rarr B

The semantic type A rarr B is a direct mapping from the components of thesyntactic type B A A is the semantic type of category A which is the type ofthe expected answer B is the semantic type of category B which is the type ofthe question-answer sequence

Notice that the type connective has an additional index We use this indexto capture a compositional difference between predicates and arguments ona sentential level (structural composition relation ) and between questionsand answers on a dialogue level (structural composition relation ) Follow-ing the structured meaning approach we assume question-answer sequencesto form a syntactic and semantic unit Syntactically we assume the question-answer sequence to belong to category s1 Semantically the question-answersentence is a proposition which has a certain truth value similar to declarativeclauses Now that we have defined the general type schema for wh-questionswe continue to illustrate how such a type for wh-questions can account fordifferent types of question-answer combinations

In most languages wh-questions can range over any type of answer de-pending on the kind of wh-phrase For instance in English a single con-stituent question with lsquowhorsquo or lsquowhatrsquo requires a referential noun phrase (Johnthe man him) or a quantified noun phrase as an answer (everyone every man)Whereas wh-questions with adjunct wh-phrases such as lsquowherersquo or lsquowhenrsquo re-quire a locative or temporal modifier as response Additionally we can alsohave multiple wh-questions that require pair-list answers As these examples

1For present purposes we treat declarative clauses and question-answer sequences as syn-tactically uniform Further research is needed to discover how the syntactic distinction betweendifferent statements in a discourse can be captured in a type-logical grammar approach

158 5 Syntax-semantics interface of wh-questions

illustrate both the answer type A and the question type B in the type defini-tion of wh-questions (see definition 25) may be complex categories As an ex-ample we list a number of possible types of wh-questions with correspondingsemantic types

lsquoWhich man rsquo snp e rarr tlsquoWho rsquo s(s(nps)) ((e rarr t) rarr t) rarr tlsquoWhen rsquo s(iviv) ((e rarr t) rarr (e rarr t)) rarr tlsquoWho whatrsquo (snp)np e rarr (e rarr t)

Sample derivation To illustrate how a wh-question combines with an ex-pression with a matching type we derive the constituent question lsquoWho sawMaryrsquo This question can be answered with a higher-order typed answer forinstance the generalized quantifier lsquoNobodyrsquo The constituent question has goaltype s(s(nps)) The required answer type is the higher-order type s(nps)(abbreviated to gq) which maps to semantic type ((e rarr t) rarr t) lsquoNobodyrsquo canbe used as an answer because it matches the required type The wh-questionselects its answer and the combination yields a question-answer sequence oftype s with semantic type t The derivation step of merging a wh-questionlsquoWho saw Maryrsquo with a possible answer lsquoNobodyrsquo is as follows

(Who (saw Mary) ` sgq Nobody ` gq(Who (saw Mary)) Nobody ` s

[E]

Now that we know the syntactic and semantic type of wh-questions let ussee how this type determines the meaning assembly of wh-questions

522 Meaning assembly of wh-questionsIn chapter 3 we introduced the semantic operator ω to represent the meaningassembly of any type of wh-question The function of ω was to bind the gaphypothesis in the semantic term of a wh-question We repeat the inferencerule along with the meaning assembly that accounted for merging a wh-typeschema with a question body2

Γ ` ω WH(A B Bprime Aprime) ∆[x A] ` t B∆[Γ] ` (ω λxt) Bprime Aprime [WH]

The inference rule shows that the merge step is both an abstraction step of thegap hypothesis as well as an application step for merging the wh-phrase to thebody of the question After merging the wh-phrase we obtain a wh-questionof type Bprime Aprime which is the decomposed type of wh-questions Notice that

2We abstract here over the different operations that have been proposed for the different ver-sions of the wh-type schemata As explained in chapter 2 structural differences are not reflectedin the semantic representation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 2: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

154 5 Syntax-semantics interface of wh-questions

wh-type schema by reanalyzing some of the cross-linguistic data from chapter4 and we show that we can account for the correct syntactic type and meaningassembly of wh-questions

51 Semantics of questions and answers

The type-logical grammar approach imposes a strong correlation between thesyntactic construction of an expression and its semantic interpretation Thiscorrelation is formulated in the Curry-Howard correspondence between syn-tax and semantics The correspondence stems from Fregersquos principle of com-positionality The meaning of the words (= lexical semantics) and the waythey are put together (= derivational semantics) determines the meaning ofthe whole

Many theories that account for the semantics of questions relate the mean-ing of a question to its possible answers (for an overview of those theoriessee Groenendijk and Stokhof (1997)) We present two approaches of relatingquestions and answers the proposition set approach (Hamblin 1958 Karttunen1977 Groenendijk and Stokhof 1984) in which questions represent propos-itions and the approach which Krifka (2001) named the structured meaningapproach also referred to as the functional or categorial approach (Groenendijkand Stokhof 1984) In this latter approach the interrogative in combinationwith its answer forms a statement

511 Proposition set approachThe proposition set approach was formulated by Hamblin (1958) who wrote aset of principles that largely influenced the logical approach to the semanticsof questions (Karttunen 1977 Groenendijk and Stokhof 1984) With theseprinciples Hamblin (1958) captures the main idea behind a logical approachto interrogatives which states that to determine the meaning of an interrogat-ive one has to inspect what kind of statement can serve as a response The firstof Hamblinrsquos principles was paraphrased by Groenendijk and Stokhof (1997)as ldquoan answer to a question is a sentence or statementrdquo The principle imple-ments the idea that the semantic status of an answer is a proposition and thatthe syntactic form of an answer is irrelevant

Answers as statements Constituent questions such as lsquoWhat did Mary readrsquoor lsquoWho read The Minimalist Programrsquo receive a full statement as an answer Alikely answer to either question could be lsquoMary read The Minimalist ProgramrsquoAnswers which only give a single constituent as an answer (lsquoThe Minimalistprogramrsquo or lsquoMaryrsquo) are said to be derived from the full statement Thus theunderlying structure of an answer has to be a complete statement with therequested argument embedded in it

51 Semantics of questions and answers 155

The semantics of questions in the proposition set approach are representedin variants of predicate logic Adopting the presentation style of Krifka (2001)we illustrate the semantics of a question lsquoWho read The Minimalist Programrsquo inpredicate logic

Question lsquoWho read the Minimalist Programrsquop | existx(person x)and p = ((read mp) x)

= ((read mp) x)|(person x))= ((read mp) m ((read mp) j

Answer lsquoJohn read the Minimalist programrsquo= ((read mp) j

The answer to such a question is represented by the set of propositions p thatare restricted by the predicates that have to hold for this proposition For ex-ample there is a person for which it is true that this person read the MinimalistProgram By analyzing the restrictions as part of the proposition the answeryields all interpretations that make such a proposition true

The main critique of the proposition set approach is that it lacks express-ibility Because Hamblin (1958) chooses to ignore the syntactic form of theanswer and treats all answers as full statements the proposition set approachcannot directly relate a certain type of answer to a certain type of questionKrifka (2001) notes that questions with alternative answers or multiple con-stituent questions cannot be handled easily in the proposition set approachThe dominant view is that questions belong to a single category and shouldtherefore be mapped to one type This stands in contrast to the view whichis reflected by the structured meaning approach which takes a polymorphicstance towards the categorial nature of questions

512 Structured meaning approachThe structured meaning approach is sometimes referred to as the functional orcategorial approach The approach is developed by logicians and semanticistsand supports the idea that the meaning of a question is dependent on themeaning of the answer and vice versa Along similar lines Hiz (1978) pointsout that questions and their answers are not autonomous sentences but thatthey form a semantic unit mdash a question-answer pair We briefly discuss thestructured meaning approach and its syntactic consequences

Constituent answers An appropriate answer to a single constituent ques-tion may be any type of syntactic object This might be a generalized quantifierphrase or a verb phrase as well as a noun phrase or prepositional phrase Ad-ditionally in multiple wh-questions different combinations of syntactic ob-jects can be used as an answer The wh-question directs the kind of answersthat can be expected

(51) a lsquoWho saw Maryrsquo John nobody Johnrsquos sister

b lsquoWhen did John see Maryrsquo In the afternoon

156 5 Syntax-semantics interface of wh-questions

c lsquoWhich man did John seersquo His father the neighbor

d lsquoWhy did John see Maryrsquo Because

e lsquoWho saw whomrsquopair list reading John (saw) Bill Mary (saw) Sue functional reading every professorhis student Johnhis sister

As the sentences illustrate the answers have a direct relation to the interrog-ative phrase in the question To capture the relation between the question andits possible answer type the structured meaning approach formulates the ideathat the question and answer form a unit both syntactically and semanticallySyntactically the interrogative in combination with its answer forms an indic-ative sentence or a question-answer sequence This syntactic unit is reflectedin the semantics where the question meaning is a function that yields a pro-position when applied to the meaning of answer (Krifka 2001)

In the proposition set approach answers belong to a uniform type of syn-tactic object viz s whereas the answers in the structured meaning approachmay fall into any kind of category A syntactic analysis must formulate a wayto account for the possible variation in question-answer pairs Along similarlines Groenendijk and Stokhof (1984) have argued that a proper theory of thesemantics of questions should not focus on one type for wh-questions butdefine a multitude of types and determine the semantic relationship betweenthem Groenendijk and Stokhof (1997) refer to this view as ldquothe polymorphicstancerdquo

Within the type-logical grammar framework a polymorphic view on ques-tion and answer types comes quite naturally as shown in work of Hausser(1983) and more recently in Bernardi and Moot (2003) These works showthat the diversity in answer types can also be derived from uniformily typedwh-phrases We will follow this line of ldquoderivational polymorphismrdquo andshow that by incorporating the answer type into the type assigned to wh-phrases we can account for different question-answer sequences on the basisof a single type for wh-questions

52 Question and answer types

In chapter 3 we assigned type wh to wh-questions as a temporary abbrevi-ation In a structured meaning approach questions are expected to be func-tions that when applied to an answer yield a proposition In this section wewill spell out the wh abbreviation in such a way that it reflects the functor-argument relation between a wh-question and its response In section 521we determine how this relation is captured in the syntactic and semantic typedefinition of wh-questions The next step is to determine the lexical semanticsof wh-phrases In section 522 we work out the lambda term for the ω oper-ator that we have used so far as a semantic operator for wh-questions In sec-tion 523 we will show that spelling out the wh-question type along with the

52 Question and answer types 157

meaning assembly for the ω-operator leads to derivability patterns betweeninstances of the wh-type schema These patterns follow from characteristictype-shifting laws that hold in the semantic type language

521 Type definition of wh-questionsSo far we have assumed wh-questions to be expressions of type wh Adopt-ing a structured meaning approach of questions we will incorporate the typeof possible answers into the type of the wh-question Generalizing over thepossible types of answers and questions we decompose the type abbreviationwh of wh-questions into the following type

Definition 25 Decomposition of wh-question type wh

syntactic type semantic type

B A = A rarr B

The semantic type A rarr B is a direct mapping from the components of thesyntactic type B A A is the semantic type of category A which is the type ofthe expected answer B is the semantic type of category B which is the type ofthe question-answer sequence

Notice that the type connective has an additional index We use this indexto capture a compositional difference between predicates and arguments ona sentential level (structural composition relation ) and between questionsand answers on a dialogue level (structural composition relation ) Follow-ing the structured meaning approach we assume question-answer sequencesto form a syntactic and semantic unit Syntactically we assume the question-answer sequence to belong to category s1 Semantically the question-answersentence is a proposition which has a certain truth value similar to declarativeclauses Now that we have defined the general type schema for wh-questionswe continue to illustrate how such a type for wh-questions can account fordifferent types of question-answer combinations

In most languages wh-questions can range over any type of answer de-pending on the kind of wh-phrase For instance in English a single con-stituent question with lsquowhorsquo or lsquowhatrsquo requires a referential noun phrase (Johnthe man him) or a quantified noun phrase as an answer (everyone every man)Whereas wh-questions with adjunct wh-phrases such as lsquowherersquo or lsquowhenrsquo re-quire a locative or temporal modifier as response Additionally we can alsohave multiple wh-questions that require pair-list answers As these examples

1For present purposes we treat declarative clauses and question-answer sequences as syn-tactically uniform Further research is needed to discover how the syntactic distinction betweendifferent statements in a discourse can be captured in a type-logical grammar approach

158 5 Syntax-semantics interface of wh-questions

illustrate both the answer type A and the question type B in the type defini-tion of wh-questions (see definition 25) may be complex categories As an ex-ample we list a number of possible types of wh-questions with correspondingsemantic types

lsquoWhich man rsquo snp e rarr tlsquoWho rsquo s(s(nps)) ((e rarr t) rarr t) rarr tlsquoWhen rsquo s(iviv) ((e rarr t) rarr (e rarr t)) rarr tlsquoWho whatrsquo (snp)np e rarr (e rarr t)

Sample derivation To illustrate how a wh-question combines with an ex-pression with a matching type we derive the constituent question lsquoWho sawMaryrsquo This question can be answered with a higher-order typed answer forinstance the generalized quantifier lsquoNobodyrsquo The constituent question has goaltype s(s(nps)) The required answer type is the higher-order type s(nps)(abbreviated to gq) which maps to semantic type ((e rarr t) rarr t) lsquoNobodyrsquo canbe used as an answer because it matches the required type The wh-questionselects its answer and the combination yields a question-answer sequence oftype s with semantic type t The derivation step of merging a wh-questionlsquoWho saw Maryrsquo with a possible answer lsquoNobodyrsquo is as follows

(Who (saw Mary) ` sgq Nobody ` gq(Who (saw Mary)) Nobody ` s

[E]

Now that we know the syntactic and semantic type of wh-questions let ussee how this type determines the meaning assembly of wh-questions

522 Meaning assembly of wh-questionsIn chapter 3 we introduced the semantic operator ω to represent the meaningassembly of any type of wh-question The function of ω was to bind the gaphypothesis in the semantic term of a wh-question We repeat the inferencerule along with the meaning assembly that accounted for merging a wh-typeschema with a question body2

Γ ` ω WH(A B Bprime Aprime) ∆[x A] ` t B∆[Γ] ` (ω λxt) Bprime Aprime [WH]

The inference rule shows that the merge step is both an abstraction step of thegap hypothesis as well as an application step for merging the wh-phrase to thebody of the question After merging the wh-phrase we obtain a wh-questionof type Bprime Aprime which is the decomposed type of wh-questions Notice that

2We abstract here over the different operations that have been proposed for the different ver-sions of the wh-type schemata As explained in chapter 2 structural differences are not reflectedin the semantic representation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 3: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

51 Semantics of questions and answers 155

The semantics of questions in the proposition set approach are representedin variants of predicate logic Adopting the presentation style of Krifka (2001)we illustrate the semantics of a question lsquoWho read The Minimalist Programrsquo inpredicate logic

Question lsquoWho read the Minimalist Programrsquop | existx(person x)and p = ((read mp) x)

= ((read mp) x)|(person x))= ((read mp) m ((read mp) j

Answer lsquoJohn read the Minimalist programrsquo= ((read mp) j

The answer to such a question is represented by the set of propositions p thatare restricted by the predicates that have to hold for this proposition For ex-ample there is a person for which it is true that this person read the MinimalistProgram By analyzing the restrictions as part of the proposition the answeryields all interpretations that make such a proposition true

The main critique of the proposition set approach is that it lacks express-ibility Because Hamblin (1958) chooses to ignore the syntactic form of theanswer and treats all answers as full statements the proposition set approachcannot directly relate a certain type of answer to a certain type of questionKrifka (2001) notes that questions with alternative answers or multiple con-stituent questions cannot be handled easily in the proposition set approachThe dominant view is that questions belong to a single category and shouldtherefore be mapped to one type This stands in contrast to the view whichis reflected by the structured meaning approach which takes a polymorphicstance towards the categorial nature of questions

512 Structured meaning approachThe structured meaning approach is sometimes referred to as the functional orcategorial approach The approach is developed by logicians and semanticistsand supports the idea that the meaning of a question is dependent on themeaning of the answer and vice versa Along similar lines Hiz (1978) pointsout that questions and their answers are not autonomous sentences but thatthey form a semantic unit mdash a question-answer pair We briefly discuss thestructured meaning approach and its syntactic consequences

Constituent answers An appropriate answer to a single constituent ques-tion may be any type of syntactic object This might be a generalized quantifierphrase or a verb phrase as well as a noun phrase or prepositional phrase Ad-ditionally in multiple wh-questions different combinations of syntactic ob-jects can be used as an answer The wh-question directs the kind of answersthat can be expected

(51) a lsquoWho saw Maryrsquo John nobody Johnrsquos sister

b lsquoWhen did John see Maryrsquo In the afternoon

156 5 Syntax-semantics interface of wh-questions

c lsquoWhich man did John seersquo His father the neighbor

d lsquoWhy did John see Maryrsquo Because

e lsquoWho saw whomrsquopair list reading John (saw) Bill Mary (saw) Sue functional reading every professorhis student Johnhis sister

As the sentences illustrate the answers have a direct relation to the interrog-ative phrase in the question To capture the relation between the question andits possible answer type the structured meaning approach formulates the ideathat the question and answer form a unit both syntactically and semanticallySyntactically the interrogative in combination with its answer forms an indic-ative sentence or a question-answer sequence This syntactic unit is reflectedin the semantics where the question meaning is a function that yields a pro-position when applied to the meaning of answer (Krifka 2001)

In the proposition set approach answers belong to a uniform type of syn-tactic object viz s whereas the answers in the structured meaning approachmay fall into any kind of category A syntactic analysis must formulate a wayto account for the possible variation in question-answer pairs Along similarlines Groenendijk and Stokhof (1984) have argued that a proper theory of thesemantics of questions should not focus on one type for wh-questions butdefine a multitude of types and determine the semantic relationship betweenthem Groenendijk and Stokhof (1997) refer to this view as ldquothe polymorphicstancerdquo

Within the type-logical grammar framework a polymorphic view on ques-tion and answer types comes quite naturally as shown in work of Hausser(1983) and more recently in Bernardi and Moot (2003) These works showthat the diversity in answer types can also be derived from uniformily typedwh-phrases We will follow this line of ldquoderivational polymorphismrdquo andshow that by incorporating the answer type into the type assigned to wh-phrases we can account for different question-answer sequences on the basisof a single type for wh-questions

52 Question and answer types

In chapter 3 we assigned type wh to wh-questions as a temporary abbrevi-ation In a structured meaning approach questions are expected to be func-tions that when applied to an answer yield a proposition In this section wewill spell out the wh abbreviation in such a way that it reflects the functor-argument relation between a wh-question and its response In section 521we determine how this relation is captured in the syntactic and semantic typedefinition of wh-questions The next step is to determine the lexical semanticsof wh-phrases In section 522 we work out the lambda term for the ω oper-ator that we have used so far as a semantic operator for wh-questions In sec-tion 523 we will show that spelling out the wh-question type along with the

52 Question and answer types 157

meaning assembly for the ω-operator leads to derivability patterns betweeninstances of the wh-type schema These patterns follow from characteristictype-shifting laws that hold in the semantic type language

521 Type definition of wh-questionsSo far we have assumed wh-questions to be expressions of type wh Adopt-ing a structured meaning approach of questions we will incorporate the typeof possible answers into the type of the wh-question Generalizing over thepossible types of answers and questions we decompose the type abbreviationwh of wh-questions into the following type

Definition 25 Decomposition of wh-question type wh

syntactic type semantic type

B A = A rarr B

The semantic type A rarr B is a direct mapping from the components of thesyntactic type B A A is the semantic type of category A which is the type ofthe expected answer B is the semantic type of category B which is the type ofthe question-answer sequence

Notice that the type connective has an additional index We use this indexto capture a compositional difference between predicates and arguments ona sentential level (structural composition relation ) and between questionsand answers on a dialogue level (structural composition relation ) Follow-ing the structured meaning approach we assume question-answer sequencesto form a syntactic and semantic unit Syntactically we assume the question-answer sequence to belong to category s1 Semantically the question-answersentence is a proposition which has a certain truth value similar to declarativeclauses Now that we have defined the general type schema for wh-questionswe continue to illustrate how such a type for wh-questions can account fordifferent types of question-answer combinations

In most languages wh-questions can range over any type of answer de-pending on the kind of wh-phrase For instance in English a single con-stituent question with lsquowhorsquo or lsquowhatrsquo requires a referential noun phrase (Johnthe man him) or a quantified noun phrase as an answer (everyone every man)Whereas wh-questions with adjunct wh-phrases such as lsquowherersquo or lsquowhenrsquo re-quire a locative or temporal modifier as response Additionally we can alsohave multiple wh-questions that require pair-list answers As these examples

1For present purposes we treat declarative clauses and question-answer sequences as syn-tactically uniform Further research is needed to discover how the syntactic distinction betweendifferent statements in a discourse can be captured in a type-logical grammar approach

158 5 Syntax-semantics interface of wh-questions

illustrate both the answer type A and the question type B in the type defini-tion of wh-questions (see definition 25) may be complex categories As an ex-ample we list a number of possible types of wh-questions with correspondingsemantic types

lsquoWhich man rsquo snp e rarr tlsquoWho rsquo s(s(nps)) ((e rarr t) rarr t) rarr tlsquoWhen rsquo s(iviv) ((e rarr t) rarr (e rarr t)) rarr tlsquoWho whatrsquo (snp)np e rarr (e rarr t)

Sample derivation To illustrate how a wh-question combines with an ex-pression with a matching type we derive the constituent question lsquoWho sawMaryrsquo This question can be answered with a higher-order typed answer forinstance the generalized quantifier lsquoNobodyrsquo The constituent question has goaltype s(s(nps)) The required answer type is the higher-order type s(nps)(abbreviated to gq) which maps to semantic type ((e rarr t) rarr t) lsquoNobodyrsquo canbe used as an answer because it matches the required type The wh-questionselects its answer and the combination yields a question-answer sequence oftype s with semantic type t The derivation step of merging a wh-questionlsquoWho saw Maryrsquo with a possible answer lsquoNobodyrsquo is as follows

(Who (saw Mary) ` sgq Nobody ` gq(Who (saw Mary)) Nobody ` s

[E]

Now that we know the syntactic and semantic type of wh-questions let ussee how this type determines the meaning assembly of wh-questions

522 Meaning assembly of wh-questionsIn chapter 3 we introduced the semantic operator ω to represent the meaningassembly of any type of wh-question The function of ω was to bind the gaphypothesis in the semantic term of a wh-question We repeat the inferencerule along with the meaning assembly that accounted for merging a wh-typeschema with a question body2

Γ ` ω WH(A B Bprime Aprime) ∆[x A] ` t B∆[Γ] ` (ω λxt) Bprime Aprime [WH]

The inference rule shows that the merge step is both an abstraction step of thegap hypothesis as well as an application step for merging the wh-phrase to thebody of the question After merging the wh-phrase we obtain a wh-questionof type Bprime Aprime which is the decomposed type of wh-questions Notice that

2We abstract here over the different operations that have been proposed for the different ver-sions of the wh-type schemata As explained in chapter 2 structural differences are not reflectedin the semantic representation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 4: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

156 5 Syntax-semantics interface of wh-questions

c lsquoWhich man did John seersquo His father the neighbor

d lsquoWhy did John see Maryrsquo Because

e lsquoWho saw whomrsquopair list reading John (saw) Bill Mary (saw) Sue functional reading every professorhis student Johnhis sister

As the sentences illustrate the answers have a direct relation to the interrog-ative phrase in the question To capture the relation between the question andits possible answer type the structured meaning approach formulates the ideathat the question and answer form a unit both syntactically and semanticallySyntactically the interrogative in combination with its answer forms an indic-ative sentence or a question-answer sequence This syntactic unit is reflectedin the semantics where the question meaning is a function that yields a pro-position when applied to the meaning of answer (Krifka 2001)

In the proposition set approach answers belong to a uniform type of syn-tactic object viz s whereas the answers in the structured meaning approachmay fall into any kind of category A syntactic analysis must formulate a wayto account for the possible variation in question-answer pairs Along similarlines Groenendijk and Stokhof (1984) have argued that a proper theory of thesemantics of questions should not focus on one type for wh-questions butdefine a multitude of types and determine the semantic relationship betweenthem Groenendijk and Stokhof (1997) refer to this view as ldquothe polymorphicstancerdquo

Within the type-logical grammar framework a polymorphic view on ques-tion and answer types comes quite naturally as shown in work of Hausser(1983) and more recently in Bernardi and Moot (2003) These works showthat the diversity in answer types can also be derived from uniformily typedwh-phrases We will follow this line of ldquoderivational polymorphismrdquo andshow that by incorporating the answer type into the type assigned to wh-phrases we can account for different question-answer sequences on the basisof a single type for wh-questions

52 Question and answer types

In chapter 3 we assigned type wh to wh-questions as a temporary abbrevi-ation In a structured meaning approach questions are expected to be func-tions that when applied to an answer yield a proposition In this section wewill spell out the wh abbreviation in such a way that it reflects the functor-argument relation between a wh-question and its response In section 521we determine how this relation is captured in the syntactic and semantic typedefinition of wh-questions The next step is to determine the lexical semanticsof wh-phrases In section 522 we work out the lambda term for the ω oper-ator that we have used so far as a semantic operator for wh-questions In sec-tion 523 we will show that spelling out the wh-question type along with the

52 Question and answer types 157

meaning assembly for the ω-operator leads to derivability patterns betweeninstances of the wh-type schema These patterns follow from characteristictype-shifting laws that hold in the semantic type language

521 Type definition of wh-questionsSo far we have assumed wh-questions to be expressions of type wh Adopt-ing a structured meaning approach of questions we will incorporate the typeof possible answers into the type of the wh-question Generalizing over thepossible types of answers and questions we decompose the type abbreviationwh of wh-questions into the following type

Definition 25 Decomposition of wh-question type wh

syntactic type semantic type

B A = A rarr B

The semantic type A rarr B is a direct mapping from the components of thesyntactic type B A A is the semantic type of category A which is the type ofthe expected answer B is the semantic type of category B which is the type ofthe question-answer sequence

Notice that the type connective has an additional index We use this indexto capture a compositional difference between predicates and arguments ona sentential level (structural composition relation ) and between questionsand answers on a dialogue level (structural composition relation ) Follow-ing the structured meaning approach we assume question-answer sequencesto form a syntactic and semantic unit Syntactically we assume the question-answer sequence to belong to category s1 Semantically the question-answersentence is a proposition which has a certain truth value similar to declarativeclauses Now that we have defined the general type schema for wh-questionswe continue to illustrate how such a type for wh-questions can account fordifferent types of question-answer combinations

In most languages wh-questions can range over any type of answer de-pending on the kind of wh-phrase For instance in English a single con-stituent question with lsquowhorsquo or lsquowhatrsquo requires a referential noun phrase (Johnthe man him) or a quantified noun phrase as an answer (everyone every man)Whereas wh-questions with adjunct wh-phrases such as lsquowherersquo or lsquowhenrsquo re-quire a locative or temporal modifier as response Additionally we can alsohave multiple wh-questions that require pair-list answers As these examples

1For present purposes we treat declarative clauses and question-answer sequences as syn-tactically uniform Further research is needed to discover how the syntactic distinction betweendifferent statements in a discourse can be captured in a type-logical grammar approach

158 5 Syntax-semantics interface of wh-questions

illustrate both the answer type A and the question type B in the type defini-tion of wh-questions (see definition 25) may be complex categories As an ex-ample we list a number of possible types of wh-questions with correspondingsemantic types

lsquoWhich man rsquo snp e rarr tlsquoWho rsquo s(s(nps)) ((e rarr t) rarr t) rarr tlsquoWhen rsquo s(iviv) ((e rarr t) rarr (e rarr t)) rarr tlsquoWho whatrsquo (snp)np e rarr (e rarr t)

Sample derivation To illustrate how a wh-question combines with an ex-pression with a matching type we derive the constituent question lsquoWho sawMaryrsquo This question can be answered with a higher-order typed answer forinstance the generalized quantifier lsquoNobodyrsquo The constituent question has goaltype s(s(nps)) The required answer type is the higher-order type s(nps)(abbreviated to gq) which maps to semantic type ((e rarr t) rarr t) lsquoNobodyrsquo canbe used as an answer because it matches the required type The wh-questionselects its answer and the combination yields a question-answer sequence oftype s with semantic type t The derivation step of merging a wh-questionlsquoWho saw Maryrsquo with a possible answer lsquoNobodyrsquo is as follows

(Who (saw Mary) ` sgq Nobody ` gq(Who (saw Mary)) Nobody ` s

[E]

Now that we know the syntactic and semantic type of wh-questions let ussee how this type determines the meaning assembly of wh-questions

522 Meaning assembly of wh-questionsIn chapter 3 we introduced the semantic operator ω to represent the meaningassembly of any type of wh-question The function of ω was to bind the gaphypothesis in the semantic term of a wh-question We repeat the inferencerule along with the meaning assembly that accounted for merging a wh-typeschema with a question body2

Γ ` ω WH(A B Bprime Aprime) ∆[x A] ` t B∆[Γ] ` (ω λxt) Bprime Aprime [WH]

The inference rule shows that the merge step is both an abstraction step of thegap hypothesis as well as an application step for merging the wh-phrase to thebody of the question After merging the wh-phrase we obtain a wh-questionof type Bprime Aprime which is the decomposed type of wh-questions Notice that

2We abstract here over the different operations that have been proposed for the different ver-sions of the wh-type schemata As explained in chapter 2 structural differences are not reflectedin the semantic representation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 5: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

52 Question and answer types 157

meaning assembly for the ω-operator leads to derivability patterns betweeninstances of the wh-type schema These patterns follow from characteristictype-shifting laws that hold in the semantic type language

521 Type definition of wh-questionsSo far we have assumed wh-questions to be expressions of type wh Adopt-ing a structured meaning approach of questions we will incorporate the typeof possible answers into the type of the wh-question Generalizing over thepossible types of answers and questions we decompose the type abbreviationwh of wh-questions into the following type

Definition 25 Decomposition of wh-question type wh

syntactic type semantic type

B A = A rarr B

The semantic type A rarr B is a direct mapping from the components of thesyntactic type B A A is the semantic type of category A which is the type ofthe expected answer B is the semantic type of category B which is the type ofthe question-answer sequence

Notice that the type connective has an additional index We use this indexto capture a compositional difference between predicates and arguments ona sentential level (structural composition relation ) and between questionsand answers on a dialogue level (structural composition relation ) Follow-ing the structured meaning approach we assume question-answer sequencesto form a syntactic and semantic unit Syntactically we assume the question-answer sequence to belong to category s1 Semantically the question-answersentence is a proposition which has a certain truth value similar to declarativeclauses Now that we have defined the general type schema for wh-questionswe continue to illustrate how such a type for wh-questions can account fordifferent types of question-answer combinations

In most languages wh-questions can range over any type of answer de-pending on the kind of wh-phrase For instance in English a single con-stituent question with lsquowhorsquo or lsquowhatrsquo requires a referential noun phrase (Johnthe man him) or a quantified noun phrase as an answer (everyone every man)Whereas wh-questions with adjunct wh-phrases such as lsquowherersquo or lsquowhenrsquo re-quire a locative or temporal modifier as response Additionally we can alsohave multiple wh-questions that require pair-list answers As these examples

1For present purposes we treat declarative clauses and question-answer sequences as syn-tactically uniform Further research is needed to discover how the syntactic distinction betweendifferent statements in a discourse can be captured in a type-logical grammar approach

158 5 Syntax-semantics interface of wh-questions

illustrate both the answer type A and the question type B in the type defini-tion of wh-questions (see definition 25) may be complex categories As an ex-ample we list a number of possible types of wh-questions with correspondingsemantic types

lsquoWhich man rsquo snp e rarr tlsquoWho rsquo s(s(nps)) ((e rarr t) rarr t) rarr tlsquoWhen rsquo s(iviv) ((e rarr t) rarr (e rarr t)) rarr tlsquoWho whatrsquo (snp)np e rarr (e rarr t)

Sample derivation To illustrate how a wh-question combines with an ex-pression with a matching type we derive the constituent question lsquoWho sawMaryrsquo This question can be answered with a higher-order typed answer forinstance the generalized quantifier lsquoNobodyrsquo The constituent question has goaltype s(s(nps)) The required answer type is the higher-order type s(nps)(abbreviated to gq) which maps to semantic type ((e rarr t) rarr t) lsquoNobodyrsquo canbe used as an answer because it matches the required type The wh-questionselects its answer and the combination yields a question-answer sequence oftype s with semantic type t The derivation step of merging a wh-questionlsquoWho saw Maryrsquo with a possible answer lsquoNobodyrsquo is as follows

(Who (saw Mary) ` sgq Nobody ` gq(Who (saw Mary)) Nobody ` s

[E]

Now that we know the syntactic and semantic type of wh-questions let ussee how this type determines the meaning assembly of wh-questions

522 Meaning assembly of wh-questionsIn chapter 3 we introduced the semantic operator ω to represent the meaningassembly of any type of wh-question The function of ω was to bind the gaphypothesis in the semantic term of a wh-question We repeat the inferencerule along with the meaning assembly that accounted for merging a wh-typeschema with a question body2

Γ ` ω WH(A B Bprime Aprime) ∆[x A] ` t B∆[Γ] ` (ω λxt) Bprime Aprime [WH]

The inference rule shows that the merge step is both an abstraction step of thegap hypothesis as well as an application step for merging the wh-phrase to thebody of the question After merging the wh-phrase we obtain a wh-questionof type Bprime Aprime which is the decomposed type of wh-questions Notice that

2We abstract here over the different operations that have been proposed for the different ver-sions of the wh-type schemata As explained in chapter 2 structural differences are not reflectedin the semantic representation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 6: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

158 5 Syntax-semantics interface of wh-questions

illustrate both the answer type A and the question type B in the type defini-tion of wh-questions (see definition 25) may be complex categories As an ex-ample we list a number of possible types of wh-questions with correspondingsemantic types

lsquoWhich man rsquo snp e rarr tlsquoWho rsquo s(s(nps)) ((e rarr t) rarr t) rarr tlsquoWhen rsquo s(iviv) ((e rarr t) rarr (e rarr t)) rarr tlsquoWho whatrsquo (snp)np e rarr (e rarr t)

Sample derivation To illustrate how a wh-question combines with an ex-pression with a matching type we derive the constituent question lsquoWho sawMaryrsquo This question can be answered with a higher-order typed answer forinstance the generalized quantifier lsquoNobodyrsquo The constituent question has goaltype s(s(nps)) The required answer type is the higher-order type s(nps)(abbreviated to gq) which maps to semantic type ((e rarr t) rarr t) lsquoNobodyrsquo canbe used as an answer because it matches the required type The wh-questionselects its answer and the combination yields a question-answer sequence oftype s with semantic type t The derivation step of merging a wh-questionlsquoWho saw Maryrsquo with a possible answer lsquoNobodyrsquo is as follows

(Who (saw Mary) ` sgq Nobody ` gq(Who (saw Mary)) Nobody ` s

[E]

Now that we know the syntactic and semantic type of wh-questions let ussee how this type determines the meaning assembly of wh-questions

522 Meaning assembly of wh-questionsIn chapter 3 we introduced the semantic operator ω to represent the meaningassembly of any type of wh-question The function of ω was to bind the gaphypothesis in the semantic term of a wh-question We repeat the inferencerule along with the meaning assembly that accounted for merging a wh-typeschema with a question body2

Γ ` ω WH(A B Bprime Aprime) ∆[x A] ` t B∆[Γ] ` (ω λxt) Bprime Aprime [WH]

The inference rule shows that the merge step is both an abstraction step of thegap hypothesis as well as an application step for merging the wh-phrase to thebody of the question After merging the wh-phrase we obtain a wh-questionof type Bprime Aprime which is the decomposed type of wh-questions Notice that

2We abstract here over the different operations that have been proposed for the different ver-sions of the wh-type schemata As explained in chapter 2 structural differences are not reflectedin the semantic representation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 7: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

52 Question and answer types 159

the gap hypothesis A and the answer type Aprime are related As we will showthe answer type is derivable from the type of the gap hypothesis (A ` Aprime)This relation between the answer type and the gap hypothesis must also beencoded in the meaning assembly of wh-questions We have used ω to capturethe meaning assembly of a wh-question Similar to the decomposition of thewh-question type we will see how the ω-operator can be captured as a lambdaterm

The precise meaning representation of a wh-question depends howeveron the kind of wh-phrase that forms a wh-question We argue that at leastfor argument wh-phrases different wh-type schema each can be derivedfrom a single wh-type schema The basic case for wh-phrases is a wh-typeschema that ranges over higher-order typed answers WH(np s sgq) Theω-operator that captures the meaning assembly of this wh-type schema canbe regarded as a logical constant The definition of the ω-operator generaliz-ing over different types of wh-phrases is as follows

Definition 26 Semantic term decomposition of wh-operator ω

ω = λPArarrBλQ(ArarrB)rarrB(Q P)

From the basic wh-type schema WH(np s sgq) we can derive instancesof the wh-type schema whose meaning assembly can be determined by fillingin the meaning assembly of ω Before we show how instances of the wh-typeschema are derived we first illustrate that the meaning assembly of the ω-operator does indeed yield the right meaning assembly for the wh-questionlsquoWho saw Maryrsquo

Sample derivation In a single constituent question lsquowhorsquo is merged with aquestion body of type s and associates with a np typed gap hypothesis Aftermerging lsquowhorsquo with the question body the wh-phrase replaces the gap hy-pothesis in the question body The sentence becomes of type s(s(nps))a sentence which is incomplete for an answer of type s(nps) For ease ofexposition we abbreviate s(nps) to gq The syntactic type and the lexicalmeaning assembly of the wh-phrase lsquowhorsquo is

who ` λP(et)λQ(et)t(Q P) WH(np s sgq)

The following derivation illustrates that this semantic term assignment tolsquowhorsquo derives the right meaning assembly for a wh-question lsquoWho saw Maryrsquo

λPλQ(Q P) who

WH(np s sgq)x np

see saw(nps)np

m marynp

saw mary ` (see m) nps[E]

x np (saw mary) ` ((see m) x) s[E]

who (saw mary) ` sgq[WH]

λQ(Q λx((see m) x))

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 8: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

160 5 Syntax-semantics interface of wh-questions

To derive a question-answer sequence the derivation continues as illus-trated in section 521 The answer lsquonobodyrsquo has the semantic term and syntactictype-assignment λPerarrtnotexist λy(P y) s(nps) We can derive the followingmeaning assembly for the question-answer sequence lsquoWho saw Mary Nobodyrsquo

Who saw Mary

λQ(Q λx((see m) x)) s(s(nps))Nobody

λPnotexist λy(P y) s(nps)Who saw Mary Nobody ` s

[E]

(λPnotexist λy(P y) λx((see m) x))lowast

β notexist λy((see m) y)

On the basis of the simplest case for argument wh-phrases we can derivedifferent instances of the wh-type schema Using the logical constant ω wecan determine how the meaning assembly is changed accordingly

523 Derivability patterns of wh-type schemataIncorporating the answer type into the wh-type schema enables us to derivedifferent instances of the wh-type schema from a single wh-type schema Forinstance we will show that we can account for multiple wh-questions Thederivation relations between different instances of wh-type schema can be de-scribed as a derivability pattern similar to the derivability pattern of unaryfeature decorated sentence types (diams2s ` s ` 2diamss)

The derivability pattern is based on characteristic laws in semantic typelanguage

5231 Semantic derivability

The derivability pattern of wh-type schemata is based on three theorems thatare derivable in semantic type language type-lifting geach and exchange Weillustrate each rule in semantic type language and present the meaning as-sembly for each type-shifting rule

[type-lifting] A ` (A rarr B) rarr Bx 7rarr λy(y x)

[geach] B rarr A ` (C rarr B) rarr (C rarr A)x 7rarr λyλz(x (y z))

[exchange] C rarr (D rarr E) ` D rarr (C rarr E)x 7rarr λzλy((x y) z)

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 9: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

52 Question and answer types 161

Using these theorems we can derive two additional laws argument loweringand dependent geach

argument lowering The type-lifting rule shows how an arbitrary type A islifted to a type (A rarr B) rarr B In the previous section we have illustratedthat type lifting alters the answer type to fit the answer type requested by thewh-question From the type-lifting rule we can derive the rule for argumentlowering Argument lowering applies to the dependent instead of the maintype Each type can be a complex type

[argument lowering] ((A rarr B) rarr B) rarr C ` A rarr Cx 7rarr λy(x λz(z y))

dependent geach The geach rule adds an additional dependent to both themain clause type A and its argument type B Again each type may be acomplex type The exchange rule captures the reordering of two dependentsFrom the combination of geach and exchange we can derive another validtype shifting rule which we will refer to as dependent geach because the geachrule applies to the dependent of a complex type The geach rule is now ap-plied to a complex type (D rarr E)rarr A The geach rule alone would change thistype into a complex type (C rarr (D rarr E)) rarr (C rarr A) Additionally we applyexchange to the result of the geach type We obtain the following type-shiftingrule

[dependent geach] (D rarr E) rarr (B rarr A) ` (D rarr (C rarr E)) rarr (B rarr (C rarr A))x 7rarr λzλyλv((x λu((z u) v)) y)

5232 Syntactic derivability

The theorems in the semantic type language reveal that under certain assump-tions a number of type alternations are also derivable in the syntactic formulalanguage Let us investigate under what assumptions argument lowering anddependent geach are derivable in the grammatical reasoning system We willillustrate the derivability of each type-shifting rule by presenting the syntacticderivation of a specific wh-type schema Along with the syntactic derivationwe can show how the meaning assembly of the wh-phrase changes using theω-operator as a logical constant

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 10: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

162 5 Syntax-semantics interface of wh-questions

Argument lowering Argument lowering is derivable in the grammaticalreasoning system without making any structural assumptions As an ex-ample we apply argument lowering to the subject wh-phrase lsquowhorsquo and showhow the meaning assembly is changed

lsquoWhorsquo is typed in the lexicon as WH(np s sgq) The wh-type schema as-sociates with an np gap hypothesis in a s-typed question body and yields awh-question of type sgq which requires an gq-typed answer The followingderivation illustrates how we derive an argument lowered answer type for thesubject wh-phrase lsquowhorsquo We show how argument lowering can be derived forthe wh-type schema of lsquowhorsquo For a clearer presentation we use the decom-posed type for the wh-type schema (see chapter 3)

who WHlempty(np s sgq) = (sgq)(nps)

For the lexical semantics of lsquowhorsquo we use the logical constant ω The hypo-theses are each decorated with term variables In the derivation each syntacticstep reveals the meaning assembly

who ` ω (sgq)(nps) [Q nps]who nps ` (ω Q) sgq

[E]

[x np] [P nps]np nps ` (P x) s

[E]

np ` λP(P x) s(nps)[I]

(who nps) np ` (ω Q) λP(P x) s[E]

who nps ` λx((ω Q) λP(P x)) snp[I]

who ` λQλx((ω Q) λP(P x)) (snp)(nps)[I]

λQλx((ω Q) λP(P x))

Using the definition of the logical constant ω we can compute the mean-ing assembly for the argument lowered type In the semantic term that iscomputed after argument lowering (step 1) we substitute the semantic termof the ω-operator of definition 26 (step 2) After several β-reductions (step 3)and a single η-reduction (step 4) we obtain a term that is equal to the identityfunction

λQλx((ω Q) λP(P x)) (1)[Def ω] λQλx((λPprimeλQprime(Qprime Pprime) Q) λP(P x)) (2)βlowast λQλx(Q x) (3)η λQQ (4)

Schematically the derivability pattern for argument lowering can be cap-tured in the following derivability relation

WH(np s sgq) ` WH(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 11: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

52 Question and answer types 163

Dependent Geach Unlike argument lowering dependent Geach is notfreely derivable in the grammatical reasoning system Moortgat (1997) showsthat the Geach law is derivable in the syntactic formula language with the ad-dition of associativity The restricted set of displacement postulates providesthe necessary restructuring mechanisms for deriving Geach

To use the postulates specifically for the derivation of Geach variants ofthe wh-type schema we must allow the postulates to reason over the struc-tural binary operator Figure 51 presents the alternations of displacementpostulates that the structural module needs in order to reason over answerhypotheses of ex-situ wh-phrases Similarly we must add alternations to thepostulates that underlie wh-in-situ phrases (see appendix A for an overviewof the in-situ postulates)

Γ[(∆1 ∆2) ∆3] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl1]Γ[∆1 i (∆2 ∆3)] ` CΓ[(∆1 i ∆2) ∆3] ` C

[Pr1]

Γ[∆2 (∆1 ∆3)] ` CΓ[∆1 (∆2 ∆3)] ` C

[Pl2]Γ[(∆1 ∆3) i ∆2] ` CΓ(∆1 i ∆2) ∆3 ` C

[Pr2]

Figure 51 Restricted set of displacement postulates where i isin empty

With this change in the structural module we can derive dependent Geachvariants for wh-ex-situ and wh-in-situ types Schematically the geach variantof a wh-type schema can be captured by the following type change

WH(A B Bprime Aprime) ` WH(A BC (BprimeC) Aprime)

The question body type B and the goal type Bprime Aprime have an additional depend-ent type C in the dependent geach type We will use this type to capture thedependency of a wh-phrase on the occurrence of another wh-phrase ie mul-tiple wh-questions We will show how we can derive the dependent geachtype syntactically for both the ex-situ and the wh-in-situ type schema

wh-ex-situ The wh-ex-situ type schema WHrex(diams2np s sgq) is a basic wh-

type schema which has the semantic term assignment captured by the logicaloperator ω We will show that we can derive a geach variant of this wh-type schema and present the meaning assembly accordingly The meaningassembly of the derived type is computed substituting the ω-operator in theterm that is derived for the geach type

Again for a clearer presentation we decompose the type into the usuallogical connectives of the type-logical grammar system Additionally we addsubscripts middot1 and middot2 to distinguish between the gq answer type (gq1) which ispart of the wh-type schema and the answer type(gq2) which is added throughthe derivation of the geach type In the derivation we abbreviate the struc-tural occurrence of the wh-type schema to lsquowhexrsquo

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 12: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

164 5 Syntax-semantics interface of wh-questions

whex WHrex(diams2np s sgq1) = (sgq1)(sdiams2np)

Figure 52 illustrates the derivation of the geach variant of the wh-ex-situtype Along with the syntactic derivation we compute the meaning assemblyfor each derivation step

[P (sgq2)diams2np] [z diams2np]

(sgq2)diams2np diams2np ` (P z) sgq2[E] [R diams2gq2]

gq2

((sgq2)diams2np diams2np) diams2gq2 ` (P z) R s[E]

((sgq2)diams2np diams2gq2) diams2np ` (P z) R s[Pr2]

((sgq2)diams2np diams2gq2) ` λz((P z) R) sdiams2np[I]

[E] whex ` ω (sgq1)(sdiams2np)whex (sgq2)diams2np diams2gq2) ` (ω λz((P z) R)) sgq1 [Q gq1]

(whex (sgq2)diams2np diams2gq2)) gq1 ` ((ω λz((P z) R)) Q) s[E]

((whex (sgq2)diams2np) gq1) diams2gq2 ` ((ω λz((P z) R)) Q) s[Pr12]

[Ilowast]whex ` ((sdiams2gq2)gq1)((sgq2)diams2np)

λPλQλR((ω λz((P z) R)) Q)

Figure 52 Derivation of geach variant of wh-ex-situ type

The result is a dependent geach variant for whex which can be rewritten as awh-type schema

((sdiams2gq2)gq1)((sgq1)diams2np)= WHr

ex(diams2np sgq2 (sdiams2gq2)gq1)

Similar to the derivation of the argument lowering type we determine themeaning assembly of the derived instance of the wh-type schema by substitut-ing the term definition of the ω-operator3 in the obtained term (step 2) Afterβ-reduction we obtain the term of the geach wh-type schema

λPλQλR((ω λz((P z) R)) Q) (1)[ω] λPλQλR((λPprimeλQprime(Qprime Pprime) λz((P z) R)) Q) (2)βlowast λPλQλR(Q λz((P z) R)) (3)

3See definition 26 on page 159

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 13: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

52 Question and answer types 165

The derivation of a geach type also extends to the left-ex-situ variant andto the argument lowered variant of both ex-situ wh-type schemata For thelowered geach type variant we compute the semantic term using the term forthe lowered wh-type schema (ωid = λQQ) We can compute the followingterm assembly for an lsquoargument lowered geachrsquo wh-ex-situ type

WHex(diams2np s snp1) ` WHex(diams2np snp2 (sdiams2np2)np1)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

This term can be further reduced by substituting the semantic term for thelowered wh-ex-situ type for ωid

λPλxλy((ωid λz((P y) x)) z)[ωid] λPλxλy((λQQλz((P y) x)) z)βlowast λPλxλy((P y) x)

wh-in-situ In a similar fashion to the wh-ex-situ types we can derive de-pendent geach types for the wh-in-situ type schema The syntactic proof forderiving a geach type for a wh-in-situ wh-phrase proceeds along similar linesas the proof for the wh-ex-situ type schema

Allowing the restricted set of structural rules underlying the wh-in-situtype schema to apply over the composition we derive the following de-rivability patterns of wh-in-situ type schema The semantic terms that arecomputed can be further reduced using term reduction

WHin(np s sgq) ` WHin(np s snp)ω 7rarr λQλx((ω Q) λP(P x))

WHin(np s snp) ` WHin(np snp (snp)np)ωid 7rarr λPλxλy((ωid λz((P y) x)) z)

WHin(np s sgq1) ` WHin(np sgq2 (sgq2)gq1)ω 7rarr λPλQλR((ω λz((P z) R)) Q)

5233 Overview

We have shown that the two theorems of argument lowering and dependent geachare syntactically derivable in the grammatical reasoning system Applying thetwo rules on different instances of the wh-type schema gives us the followingderivability patterns between instances of wh-type schema In figure 53 the

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 14: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

166 5 Syntax-semantics interface of wh-questions

Argument lowering Dependent geach

I WH(diams2np s snp)

HWH(diams2np s sgq) WH(diams2np snp (snpdiams2np)

I WH(diams2np sgq (sgqdiams2gq)

N

Dependent geach Argument lowering

Figure 53 Derivability pattern of wh-ex-situ schema

Argument lowering Dependent geach

I WH(np s snp)

HWH(np s sgq) WH(np snp (snpnp)

I WH(np sgq (sgqgq)

N

Dependent geach Argument lowering

Figure 54 Derivability pattern of wh-in-situ schema

Argument lowering Dependent geach

I λQQ

HλPprimeλQprime(Qprime Pprime) λPprimeλxλy((Pprime y) x)

I λPλQλR(Q λz((P z) R))

N

Dependent geach Argument lowering

Figure 55 Meaning assembly of derivability patterns

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 15: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 167

derivability pattern of wh-ex-situ types is presented and in figure 54 the pat-tern of wh-in-situ types For both ex-situ and in-situ type schemata we havethe meaning assembly patterns as presented in figure 55

These derivability patterns will be used to lexically identify wh-phrasesBy assigning a wh-type schema to a wh-phrase that fits the syntactic and se-mantic properties of the wh-phrase we can use the instances that can be de-rived from that wh-type schema In the coming section we will illustrate howthe derivability pattern for the analysis of wh-question formation can be usedWe show that we can account for different kinds of answers or prevent certaintype of answers by lexically identifying the minimal type-assignment of a wh-phrase We reanalyze some phenomena that have been addressed in chapter4 and show that on the basis of the proposed decomposition of wh-questiontypes we can provide a uniform meaning assembly of wh-questions

53 Linguistic application

The syntactic decomposition of wh-question types into types that are part ofan question-answer sequence adds polymorphism to the wh-type schemataThe semantic representation of wh-questions reflects the questionrsquos require-ment for certain types of answers We have shown that two theorems ofargument lowering and dependent geach leads to a derivability pattern ofwh-type schema In this section we continue to explore the syntactic andsemantic aspects of this polymorphism for wh-question formation The cross-linguistic data that has been explored in chapter 4 provides support for ourhypothesis that the wh-type schema accounts for a uniform interpretation ofwh-questions but also that a polymorphic stance is needed to account forvariation in question-answer sequences We show that the derivational pos-sibilities of the types largely depend on the right encoding of the wh-typeschema for wh-phrases

In section 531 we focus on the derivation of single constituent questionsin English We discuss the syntactic and semantic consequences of argumentlowering for the derivation of question-answer sequences in local and non-local wh-questions In section 532 we discuss multiple wh-questons in Eng-lish Serbo-Croatian and Japanese We show that we can account for the deri-vation of multiple wh-questions on the basis of deriving geach types for bothex-situ and in-situ type schema And as a result derive the correct meaningassembly of multiple wh-questions Additionally in section 533 and 534we discuss the semantic representation of two special instances of wh-typeschema that were syntactically explored in chapter 4 In section 533 we ex-plore pied-piping constructions in English and show that the analysis appliessimilarly to complex NP constructions in Japanese In section 534 we discussscope marking constructions and show that the decomposition of wh-questiontypes gives a direct question interpretation to such questions

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 16: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

168 5 Syntax-semantics interface of wh-questions

531 Single constituent questionsA single constituent question requires a single constituent answer We con-centrate here on argument wh-phrases to illustrate the relation between a wh-question and possible answers We first look at direct questions where theassociated gap hypothesis appears in the local domain We will furthermore il-lustrate the contrast between wh-pronous and wh-determiners Subsequentlywe discuss the meaning assembly of indirect wh-questions Lastly we brieflyshow how the meaning assembly of non-local wh-questions does not differfrom the meaning assembly of local wh-questions

5311 Direct questions

In a direct question in English a fronted wh-phrase associates with a np gaphypothesis The expected answer however differs on the wh-phrase Wh-questions with argument wh-phrases lsquowhatrsquo or lsquowhorsquo expect either a referen-tial or a quantified noun phrase Wh-questions with which-determiners onlyexpect a referential noun phrase as an answer On the basis of the derivab-ility pattern of wh-ex-situ types we can account for the distinction betweenthe two types of wh-phrases First we discuss the lexical type-assignment ofwh-pronouns Then we present the contrast with wh-determiners

Wh-pronouns A suitable answer to a wh-question such as lsquoWho saw Maryrsquomight be a referential noun phrase eg lsquoJohnrsquo as well as a generalized quan-tifier phrase eg lsquoeveryonersquo To allow both types of answers lsquowhorsquo and lsquowhomrsquoare assigned the following wh-type schema in the lexicon

who λPetλQ(et)t(Q P) WHlempty(np s s(s(nps)))

whom λPetλQ(et)t(Q P) WHrex(diams2np q s(s(nps)))

We use the following lexical entries to illustrate the syntactic derivationand meaning assembly of direct questions such as lsquoWho saw Maryrsquo andlsquoWhom did John seersquo

john mary jm npsaw λxλy((see x) y) (nps)npsee λy(see y) infnpdid λP(π2P π1P) q(np bull inf )every man λPforall λy(man y) rarr (P y)) s(nps)some woman λPexist λx((woman x)and (P x)) s(nps)

In section 521 we showed that a generalized noun phrase matches therequired type of the wh-question directly For a definite noun phrase withcategory np to be used as an answer the type needs to be lifted to s(nps) Weargued that type-lifting can be derived in the grammatical reasoning systemwithout making any further structural assumptions

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 17: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 169

The following derivation illustrates how a higher-order type for the nounphrase John can be derived The derived type matches the required type of thewh-question and this is also reflected in the meaning assembly of the question-answer sequence

Who saw Mary

λR(R λx((see m) x)) sgq

Johnj npe [Qet nps]

John nps ` (Q j) s[E]

John ` λQet(Q j) s(nps)[I]

Who saw Mary John ` s[E]

(λQ(Q j) λx((see m) x))lowast

β ((see m) j)

The sentences in example 52 and 53 present an overview of the differ-ent kinds of question-answer sequences that can be derived using the giventype-assignments for wh-pronouns The type that is derived for subject ornon-subject wh-questions is a s-typed clause which is incomplete for a higher-order typed np (s(nps)) A generalized quantifier phrase can be mergeddirectly while referential noun phrases such as lsquoJohnrsquo and lsquoMaryrsquo in example52b and 53b have to be lifted before they can be merged Along with the syn-tactic type lifting alters the semantic type of the answer in such a way that thelifted type matches the semantic type requested by the interrogative clauseThe semantic term is computed as usual

(52) Who saw Mary ` λQ(et)t(Q λx((see m) x)) s(s(nps)

a Answer lsquoevery manrsquo ` λPforall λy((man y) rarr (P y)) gqMeaning assembly forall λy((man y) rarr ((see m)y))

b Answer lsquoJohnrsquo ` j npλPet(P j) s(nps)

Meaning assembly (λP(P j) λx((see m) x))β (λx((see m) x) j)β ((see m) j)

(53) Who(m) did John see ` λQ(et)t(Q λy((see y) j)) s(s(nps))

a Answer lsquosome womanrsquo ` λPexist λx((woman x)and (P x)) gqMeaning assembly exist λx((woman x)and ((see x) j))

b Answer lsquoMary rsquo ` m npλPet(P m) s(nps)

Meaning assembly (λP(P m) λy((see y) j))lowast

β ((see m) j)

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 18: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

170 5 Syntax-semantics interface of wh-questions

Wh-determiners Suitable answers to wh-questions that are built with wh-determiners like lsquowhichrsquo are restricted to definite noun phrases The semanticdifference between wh-phrases and wh-determiners lies in the specific de-notation of the which-phrases As Pesetsky (1987) notes in his paper lsquowhichphrasesrsquo are Discourse linked while lsquowhorsquo or lsquowhatrsquo are generally not D-linkedAnswers to D-linked phrases are limited to np-typed phrases For instancethe wh-question lsquoWhich man saw Maryrsquo can be paraphrased as lsquoWho is the manthat saw Maryrsquo The person who utters the question and the hearer alreadyhave the background knowledge that the person who saw Mary is a manA definite answer is the only possible response This gives us evidence to as-sume that a wh-determiner has a minimal type assignment such that it derivesa question of type snp On the basis of this assumption wh-determinersare assigned the following wh-ex-situ types along with its lexical term assign-ment

which ` λVλPλx(x = ιy((V y)and (P y))) WH(npA snp)nwhere A isin s q

On the basis of this type-assignment we can derive the following question-answer sequence After merging the wh-question with its answer we derivea question-answer statement of type s which has a corresponding semanticterm of a proposition

Which man saw Mary

λx(x = ιy(man y)and ((see m) y) snpJohnj np

Which man saw Mary John ` s[E]

j = ιy((man y)and ((see m) y))

Quantified noun phrases such as lsquoeveryonersquo and lsquosome womanrsquo cannot beused as answers to D-linked wh-phrases These answers are ruled out onthe basis of the type-assignment for wh-determiners The question-answersequences in example 54a and 54b are derived using the type-assignment oflsquowhichrsquo above A higher-order type answer as given in example 54c cannot bederived

(54) a Which man saw Mary ` λx(x = ιy(man y)and ((see m) y)) snpAnswer lsquoJohnrsquo ` j npMeaning assembly j = ιy((man y)and ((see m) y))

b Which woman did John see` λx(x = ιy(woman y)and ((see y) j)) snpAnswer lsquoMary rsquo ` m npMeaning assembly m = ιy((woman y)and ((see y) j))

c Which man saw Mary ` snpAnswer lowast lsquoevery manrsquo ` λPforall λy(man y) rarr (P y)) gq

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 19: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 171

Further restrictions on possible question-answer combinations for singleconstituent questions can be finetuned using the derivability pattern of syn-tactic types diams2A ` A2diamsA (Bernardi 2002) On the basis of this derivabilityschema Bernardi accounts for the interaction and the ordering between dif-ferent types of quantified noun phrases Along the same lines the interactionbetween question-answer pairs can be studied further

5312 Indirect wh-questions

The decomposition of main clause wh-question types applies similarly to em-bedded wh-questions Main clause types have been typed as s q and whwhile embedded clauses are typed as sprime qprime and whprime respectively (see chapter3 section 33) The decomposition of wh-question types into a complex typeAB where A is the type of the question-answer combination and B is thetype of the answer will also be applied to embedded question types Schem-atically embedded interrogatives are typed as AprimeB where Aprime can be sprime sprimegqetcetera

Verbs such as lsquoknowrsquo and lsquoaskrsquo select for an embedded interrogative Unlikemain clause questions the intrinsic meaning of an embedded interrogative isthat a referring noun phase is expected to fill the argument role in the embed-ded sentence A sentence such as lsquoJohn knows who leftrsquo can never generate ameaning where the embedded wh-phrase can be filled by a quantifier phraseIt must be the selectional requirements of the interrogative verbs that imposesconstraints on the type of the embedded interrogative As an example wepresent the type-assignment of lsquoaskrsquo and lsquowhorsquo We abbreviate the higher-ordertype s(nps) to gq

ask λPλx((ask P) x) ((nps)(sprimenp))who λPλQ(Q P) WHl

empty(np sprime sprimegq)

As a result of this type-assignment the embedded interrogative must firstundergo argument lowering before being merged with the interrogative verbWe illustrate the analysis of the sentence lsquoMary asked who leftrsquo in figure 56 Inthe lambda term for the complete sentence the embedded interrogative is anargument to the predicate lsquoaskrsquo

5313 Non-local wh-questions

For non-local wh-questions where a wh-phrase associates with a gap hypo-thesis in an embedded clause the analysis of the meaning assembly is similarto the analysis of local questions (see chapter 3 section 333 for the syntacticderivation of non-local wh-questions) The syntactic restrictions on long-distance wh-questions do not effect the meaning assembly We derive thefollowing question-answer sequences using the decomposed types for wh-phrases

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 20: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

172 5 Syntax-semantics interface of wh-questions

marynp

asked(nps)(sprimenp)λPλx((ask P) x)

who

WHlempty(np s sprimegq)

leftnps

who left ` sprimegq[WHl

empty]

λQ(Q left)who left ` sprimenp

[lowering]

λy(left y)asked (who left) ` nps

[E]

mary (asked (who left)) ` s[E]

((ask λy(left y)) m)

Figure 56 Natural deduction derivation with meaning assembly of lsquoMaryasked who leftrsquo

wh-phrases WHrex(diams2np q sgq)

(55) Who did Sue believe saw Mary ` sgq

a Answer John ` np

Meaning assembly ((believe ((see m) j)) s)

b Answer Every man ` gq

Meaning assembly forall λy((man y) rarr ((believe ((see m) y)) s))

(56) Who did Sue believe John saw ` sgq

a Answer Mary ` s

Meaning assembly ((believe ((see m) j)) s)

b Answer some woman ` gq

Meaning assembly exist λx((woman x)and ((believe ((see x) j)) s))

532 Multiple wh-questionsWith the derivability pattern of wh-type schema using dependent Geach aspresented in section 523 we can derive multiple wh-questions from a single

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 21: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 173

type-assignment of the wh-type schema to a wh-phrase in the lexicon Us-ing the derivability relations between wh-ex-situ and wh-in-situ type schemawe can account for the syntactic differences between multiple wh-questions inEnglish Serbo-Croatian and Japanese English and Serbo-Croatian are wh-ex-situ languages Nevertheless English only allows one wh-phrase to be fron-ted whereas Serbo-Croatian allows multiple wh-fronting In Serbo-Croatianthe derivation of multiple wh-questions follows from the derivability of a de-pendent Geach type for wh-ex-situ type schema Additionally we discusshow the restrictions on possible answers to multiple wh-questions in Englishcan be accounted for Lastly we show that Japanese multiple wh-questionscan be derived from the derivability pattern of wh-in-situ types

5321 Multiple wh-fronting

In chapter 4 section 43 we presented data on multiple wh-questions in Serbo-Croatian In Serbo-Croatian all wh-phrases occur fronted We illustrate thatwe can derive multiple fronting in Serbo-Croatian from a single lexical type-assignment to wh-phrases based on the derivability pattern of wh-ex-situtype The free order of fronted wh-phrases is derived because we may usea geached type for either wh-phrase The wh-phrase with the geach type vari-ant is the leftmost wh-phrase We limit the analysis to the derivation of thedirect question lsquoko koga vidirsquo (= who whom sees)

Wh-phrases lsquokorsquo (= who) and lsquokogarsquo (= whom) are assigned the followingcategory and semantic term in the lexicon

ko λPetλxe(P x) WHlex(diams2NOM s sNOM) (= lsquowhorsquo)

koga λQetλye(Q y) WHrex(diams2ACC s sACC) (= lsquowhomrsquo)

For the sake of simplicity we choose to assign the wh-phrases an argumentlowered wh-ex-situ type For a proper treatment of Serbo-Croatian we shouldtreat the wh-phrases along similar lines as English where we derived wh-questions allowing a generalized quantified noun phrase as an answer

Using the above type assignments we can derive the following multiplewh-questions where either lsquokorsquo precedes lsquokogarsquo as in example 57a or wherelsquokogarsquo precedes lsquokorsquo as in example 57b

(57) a Kowho

kogawhom

vidisees

lsquoWho sees whomrsquo

b Kogawhom

kowho

vidisees

lsquoWhom was seen by whorsquo

With the use of the dependent geach type we can nog derive multiplewh-questions in Serbo-Croatian from the above single type-assignments To

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 22: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

174 5 Syntax-semantics interface of wh-questions

derive the multiple wh-question lsquoKo koga vidirsquo the fronted wh-phrase lsquokorsquo isfirst derived as a dependent geach type (see section 522 for the syntactic de-rivation) Along with the syntactic type change the lambda term of lsquokorsquo isalso changed The meaning assembly reflects the dependency of lsquokorsquo on theoccurrence of another wh-phrase

ko λRλxλy((R y) x) WHlex(diams2NOM sACC (sdiams2ACC)NOM)

We illustrate the use of the geach type by deriving lsquoko koga vidirsquo (= whowhom sees) We divide the derivation in two parts The first part of the ana-lysis shows the derivation of the question body with two gap hypotheses forboth the nominative and the accusative argument lsquoKogarsquo merges with thewh-question body and replaces the accusative gap hypothesis The resultingstructure which still contains a gap hypothesis for the nominative argumentphrase is of type sACC the category of single wh-questions

koga ` WHrex(diams2ACC s sACC)λQλy(Q y)

diams2NOM (vidi diams2ACC) ` s((see v) u)

koga (diams2NOM vidi) ` sACC[WHr

ex]

λy((see y) u)

The second part of the analysis is where the geach wh-type of the nomina-tive wh-phrase lsquokorsquo merges with the previously derived question body Thepartial structure (koga (diams2NOM vidi)) contains the right gap hypothesisdiams2NOM and has the right type for the geach type variant of lsquokorsquo

ko WHrex(diams2NOM sACC (sNOM)diams2ACC)

lsquoKorsquo merges with the structure and yields a multiple wh-question type oftype (sdiams2ACC)NOM In the derivation we omit the geach type of lsquokorsquo tofocus on the semantic representation of the multiple wh-question

koλRλxλy((R y) x)

koga (diams2NOM vidi) ` sACC

λy((see y) u)ko (koga vidi) ` (sdiams2ACC)NOM

[WHlex]

λxλy((see y) x)

The syntactic type derived for the wh-question reveals that the questionrequires two noun phrases a nominative noun phrase and an accusative nounphrase As we explained in section 523 in order to derive the type requiredfor multiple wh-questions one needs access to the displacement postulatesTherefore the answer type in the geach type for lsquokorsquo is decorated with featuresdiams2ACC

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 23: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 175

Due to the feature decoration on the added dependent we can apply thedisplacement rule to this result and derive the following argument switchingThis switching is not only structurally derived but also changes the under-lying syntactic ordering which in turn has semantic consequences The fol-lowing derivation illustrates how the ordering of the arguments for a multiplewh-phrase is derived

λyλx((see y) x))ko (koga vidi) ` (sNOM)diams2ACC [diams2ACC]

ko (koga vidi) diams2ACC ` sNOM[E]

[nom](ko (koga vidi) diams2ACC) nom ` s

[E]

(ko (koga vidi) nom) diams2ACC ` s[Pr2]

ko (koga vidi) nom ` sdiams2ACC[I]

ko (koga vidi) ` (sdiams2ACC)NOM[I]

λxλy((see y) x))

The result of this argument switching is that in possible responses to amultiple wh-question in Serbo-Croatian the argument order is free An an-swer to the question lsquoko koga vidirsquo in a context where lsquoBorisrsquo saw lsquoIvanarsquo thetwo answers lsquoBoris Ivanarsquo and lsquoIvana Borisrsquo are equally acceptable

This leads to a further speculation on the status of this feature decorationWe could interpret the additional feature information as a focus marker Thiswould be in line with Boskovic (1998) who claims that the wh-phrase thatfollows the first fronted wh-phrase is moved for focus reasons We leave it forfurther research to check whether these constructions indeed give rise to suchinterpretations

5322 Simple wh-fronting

Multiple wh-questions in English are recognized by a single wh-phrase thatappears fronted at the main clause whereas the other wh-phrases appear in-situ The ordering between wh-phrases follows a strict pattern In chapter4 we discussed how we can account for the strict ordering between wh-phrases using the derivability pattern of feature decorated s-types to dis-tinguish between different wh-phrases We now want to focus on the useof the derivability patterns between wh-type schema to derive multiple wh-questions in English

In chapter 3 section 335 we explored the syntax of multiple wh-phrasesWh-phrases that occur in-situ have been typed as WHin(npwhwh) This en-coded that the phrase may only appear in-situ in a wh-question body of typewh On the basis of the decomposed type for wh-questions the type for wh-in-situ phrases changes into the type-assignment below Notice that this typeoccurs in the derivability patterns of wh-in-situ type schema and is derivedfrom argument lowering and dependent geach

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 24: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

176 5 Syntax-semantics interface of wh-questions

wh-in-situ WHin(np snp (snp)np)

The wh-type schema encodes that a wh-phrase merges with a questionbody of type snp which contains a gap hypothesis of type np This in-stance of the wh-in-situ type schema is derived from WHin(np s sgq) usingargument lowering and dependent geach By assigning wh-in-situ phrasesthe above minimal type we correctly derive that lsquowhomrsquo can never be of typeWHin(np s sgq) In English a wh-phrase does not occur in-situ in a s-typedbody4 With this minimal type-assignment the wh-in-situ phrase is alwaysdependent on the occurrence of another wh-phrase (sgq) Nevertheless wecan add another dependent to the question body and derive a wh-in-situ typephrase which can be used in multiple wh-questions with more than two wh-phrases such as lsquoWho gave what to whomrsquo

multiple wh-in-situ WHin(np (snp)np ((snp)np)np)

The semantic term of the wh-in-situ type reflects this dependency on an-other wh-phrase In section 522 we computed the following meaning as-sembly for a lowered geach type

λP(erarr(erarrt))λxeλye((P y) x)) WHin(np snp (snp)np)

The order in which the answer types are expected is encoded both in thesyntactic type as well as in the semantic term Syntactically the wh-in-situphrase is dependent on the occurrence of the subject wh-phrase Semantic-ally the lambda abstraction binds the type of the subject wh-phrase over theobject wh-phrase On the basis of this type-assignment and the usual wh-type schema assigned to the subject wh-phrase we derive the multiple wh-question lsquoWho saw whomrsquo

who λRλQ(Q R) WHlempty(np s sgq)

4Echo-questions form an exception to the occurrence of a wh-in-situ in a s typed questionbody Echo-questions question that repeat a previously uttered sentence of which a part is notunderstood should get a different analysis than wh-questions We leave this analysis for futureresearch

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 25: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 177

whomWHin(np snp (snp)np)

λPλxλy((P y) x)

whoWH(np s sgq)

saw(nps)np [z np]

saw np ` nps[E]

who (saw np) ` sgq[WHl

empty]

λQ(Q λu((see z) u))who (saw np) ` snp

[lowering]

λu((see z) u)who (saw whom) ` (snp)np

[WHin]

λxλy((λzλu((see z) u) y) x)lowast

β λxλy((see y) x)

Possible answers to a multiple wh-questions with wh-pronouns are lim-ited to referential noun phrases By assigning wh-phrases that occur in mul-tiple wh-questions the wh-type schema that we have presented here we canaccount for these types of answers The type of generalized quantifier nounphrases do not match with the type required by the multiple wh-questionThe following examples illustrate the correct derivation of two referentialnoun phrases as answers and the type mismatches with generalized quan-tified noun phrases as answer

(58) Who saw whom ` λxλy((see y) x))) (snp)np

a John (np) Mary (np) ` ((see m) j) s

b lowast Every man (gq) some woman (gq) 0 s

c lowast John (np) some woman (gq) 0 s

d lowast Every man (gq) Mary (np) 0 s

We have shown that assigning wh-phrases a wh-in-situ type correctly ac-counts for the analysis of multiple wh-questions in English The questionwhich now arises is why does English allows only simple wh-fronting whilewe can derive multiple wh-fronting in Serbo-Croatian On the basis of the de-rivability pattern of wh-ex-situ types one could assume that similar to Serbo-Croatian it should be possible to derive a dependent geach type for wh-ex-situtypes in English Nothing prevents the derivation of such types but becauseof do-support these types are not applicable

We briefly point out why geach variants of the wh-ex-situ types in Englishcannot be used to derive multiple wh-fronting Non-subject argument wh-phrases have been assigned a wh-ex-situ type that is used in a q-typed ques-tion body to form a single constituent questions eg lsquoWhom did John seersquoIn theory we can derive a geach variant of this type which would yield thefollowing type

whom WHrex(diams2np q sgq) ` WHr

ex(diams2np qgq (sdiams2gq)gq)

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 26: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

178 5 Syntax-semantics interface of wh-questions

The geach variant is a wh-ex-situ type which can only be merged with a ques-tion body of type qgq Because question-answer combinations have beentyped as s-typed sentences the geached variant cannot be merged with anyderived wh-question As Serbo-Croatian does not have do-support and thequestion body of all wh-phrases is based on a s type sequence we can derivemultiple wh-fronting Further research should be done to see whether thisline of reasoning also extends to languages with verb second phenomena

5323 Multiple wh-in-situ

In this section we show that the geach type for wh-in-situ type schema alsoaccounts for the analysis of multiple wh-questions in a wh-in-situ languagesuch as Japanese (chapter 4 section 42) In chapter 4 we noted that the choicefor the syntactic types are led by the interpretation that a wh-question may getWe derive the same semantic representation of multiple wh-questions withwh-in-situ types as multiple wh-questions in a wh-ex-situ language (cf Serbo-Croatian)

In chapter 4 we presented data from Japanese that explored the differentkinds of wh-questions Recall that Japanese wh-questions have the same argu-ment ordering as declarative clauses The wh-phrase stays in-situ The clausewhere the wh-phrase is embedded in is marked by a question marker ie lsquokarsquoThe position of the question marker determines which clause is interpreted asa question Sentences with just a question marker and no wh-phrase get theinterpretation of a polar question We refer to chapter 4 for an overview ofexamples of Japanese wh-questions

Before we treat multiple wh-questions we illustrate that the meaningof wh-questions on the basis of wh-in-situ type schema is the same as themeaning assembly of wh-ex-situ wh-phrases In Japanese the type for singleconsituent wh-questions is qnp Based on the decomposed type for wh-questions wh-in-situ wh-phrases in Japanese are assigned wh-type schemaWHin(np q qnp) where q is the type assigned to the body of the questionheaded by the question marker lsquokarsquo5 An expression of type q denotes a yes-noquestion We leave the semantic representation of yes-no questions for fur-ther research and concentrate on the use of the q type for the derivation ofwh-questions Again we abstract away from the assignment of gq-typed an-swers and use the lower typed np-phrases as required answer types We listthe lexical entries that we have used in chapter 4 to account for the derivationof wh-questions along with their lambda term assignments We assume thewh-phrases to appear fully inflected in the lexicon

5The wh-phrase does not carry case feature information over to the answer type In Japaneseit is possible to give a single np-typed answer with the sentence-ending particle lsquodesursquo where theanswer has no case marking For instance the question lsquoJohn-wa nani-o tabe-masi-ta karsquo (= What didJohn eat) can be answered with lsquoninzin desursquo (= carrots (it is)) (Nishigauchi 1990 ex71p49)

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 27: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 179

nani dare λPP WHin(np q qnp)dare-ga λPλx(P x) WHin(NOM q qNOM)ka λQQ sqJohn Mary hon jmbook np-ga -o -ni λRR npNOMACCDATkatta λyλx((buy y) x) ACC(NOMs)

The meaning assembly of a single constituent wh-question is computed onthe basis of the following derivation and the above lexical term assignments

naniWHin(np q qnp)

mary ganom

[np]o

npaccnp o ` acc

[E] kaimasitaacc(noms)

(np o) kaimasita ` noms[E]

(mary ga) ((np o) kaimasita) ` s[E] ka

sq((mary ga) ((np o) kaimasita)) ka ` q

[E]

((mary ga) ((nani o) kaimasita)) ka ` qnp[WHin]

λx((buy x) m)

The wh-phrase lsquonanirsquo is inserted after the question marker lsquokarsquo is mergedwith the structure The meaning assembly of merging the wh-in-situ typeschema yields a semantic representation which reflects the binding of the ob-ject argument variable of the predicate lsquobuyrsquo

Let us now turn to multiple wh-questions and show that the geach typevariants of wh-in-situ type schema yields the correct meaning assembly Thefollowing examples illustrate multiple wh-questions in Japanese

(59) a dare-gawho-[nom]

nani-owhat-[acc]

kattabuy[past]

kaQ

lsquoWho ate whatrsquo

b John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

ka]Q

tazunetaasked

lsquoJohn asked who bought whatrsquoNOT lsquoWho did John ask bought whatrsquo

c lsquokarsquo marks an embedded wh-question clause

John-waJohn-[top]

[dare-gawho-[nom]

nani-owhat-[acc]

kattabought

to][Comp]

ittasaid

kaQ

lsquoWho did John say bought whatrsquo

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 28: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

180 5 Syntax-semantics interface of wh-questions

Along the same lines as for Serbo-Croatian the lsquodependent geachrsquo rule ap-plies to either of the two embedded wh-phrases The wh-phrase for whichthe geach wh-in-situ type is derived becomes dependent on the occurrence ofthe other wh-phrase and merges with the question body after the other wh-phrase is inserted Using these geach types we derive sentences with multipleembedded wh-phrases Similar to single constituent questions the embed-ded wh-phrases are interpreted at the level where the closest question markerappears

The geach type is either used for the subject wh-phrase causing the subjectwh-phrase to take scope over the object wh-phrase or vice versa

nani λQλyλx((Q y) x) WHin(np qnp (qnp)np)dare λPλxλy((P y) x) WHin(np qnp (qnp)np)

With the geach type assigned to lsquodarersquo we can derive the multiple wh-question of example 59a where lsquodarersquo is merged after lsquonani-orsquo is merged withthe question body

[np]ga

npNOM

np ga ` NOM[E]

[np]o

npACC

np o ` np[E] katta

ACC(NOMs)(np o) katta ` NOMs

[E]

(np ga) ((np o) katta) ` s[E] ka

sq((np ga) ((np o) katta)) ka ` q

[E][WHin] nani ` WHin(np q qnp)

((np ga) ((nani o) katta)) ka ` qACC[WHin] dare ` WHin(np qnp (qnp)np)

((dare ga) ((nani o) katta)) ka ` qnp

λxλy((buy y) x)

Similarly we can derive the wh-question where lsquonanirsquo is merged afterlsquodarersquo The order in which the answers are expected reflects this differenceThis analysis predicts that the question can generate two readings one wherethe subject argument is questioned with respect to the object argument andanother where we have the reverse In Japanese focus is needed to disambig-uate the wh-question

Additionally we can compute the following meaning assemblies for thewh-question in example 59 (see appendix B27 for on-line derivation)

(510) a Dare-ga nani-o katta ka ` (qnp)np

λxλy((buy y) x) (= Who bought what)

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 29: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 181

b Dare-ga nani-o katta ka ` (qnp)np

λyλx((buy y) x) (= What did who buy)

c John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λxλy((say ((buy y) x)) j) (= Who did John say bought what)

d John-wa [dare-ga nani-o katta to] itta ka ` (qnp)np

λyλx((say ((buy y) x)) j) (= What did John say who bought)

With the derivability patterns between ex-situ and wh-in-situ typeschemata we have shown that we can account for syntactic and semantic dif-ferences in question-answer pairs in single and multiple wh-question form-ation on the basis of theis polymorhism We have shown that multiple wh-question formation in multiple wh-fronting languages and wh-in-situ lan-guages can be derived from a single type-assignment for wh-phrases Ad-ditionally we have shown that in English we must identify the lexical type-assignment for wh-in-situ phrases In the coming section we broaden theidea of polymorphism on wh-type schema further and move to the analysisof pied-piping and scope marking constructions The proposed syntactic ana-lyses in chapter 4 are refined by incorporating the decomposed type for wh-questions into the suggested wh-type schema As a result we can computethe correct meaning assembly for these phenomena

533 Pied-pipingIn chapter 3 we presented a syntactic analysis of pied-piping for EnglishPied-piping is the phenomenon where the wh-phrase drags along additionalmaterial Consider the following examples of pied-piping Example 511a il-lustrates that the wh-determiner is part of a possessor phrase and pied-pipesthe whole noun phrase that is being modified Example 511b illustrates thatthe prepositional phrase can pied-pipe along with the wh-determiner Ex-ample 511c illustrates the pied piping of a complex np6

(511) a Which manrsquos picture did John see

b On which topic did John read a book

c The author of which novel did John like

To account for these questions we have presented an analysis where the wh-determiner is instantiated with the following wh-type schema

6The pied-piping of a complex noun phrase is judged as marginally acceptable Most speakersonly accept these sentences under an echo interpretation with stress intonation on lsquowhichrsquo Never-theless we incorporate this sentence into the analysis of pied-piping constructions as there are notheoretical groundson which this sentence can be ruled out in comparison to other pied-pipinganalysis Furthermore there seems an additional contrast with clearly unacceptable direct ques-tions such as lsquolowast The mother of whom did John meetrsquo This suggests a further distinction betweenwh-pronouns and wh-determiners

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 30: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

182 5 Syntax-semantics interface of wh-questions

which WHin(npAWHrex(diams2A qwh))n

where A isin np pp

Before we continue with the meaning assembly of this wh-type schemawe first paraphrase and recapitulate the use of this wh-type schema The wh-phrase embedded in a noun phrase or prepositional phrase is a wh-in-situtype which fills the position of an np-argument After the wh-phrase mergeswith the np or pp and replaces the np or pp gap hypothesis the whole con-struction becomes a wh-ex-situ type schema Thus the whole constructionbecomes like a wh-phrase that merges with a q-typed question body and re-places the np or pp gap hypothesis in the question body In this analysisadditional material is not dragged along with the wh-phrase but the wholeconstituent functions as a wh-phrase

In chapter 3 we limited the presentation of pied-piping construction to asyntactic analysis Here we will show how the meaning assembly of suchconstructions is computed For the meaning assembly we follow Morrillrsquos(1994) analysis of pied-piping in relative clause constructions To illustratethe close resemblance between pied-piping constructions in relative clausesand wh-questions we first present the meaning assembly for pied-piping inrelative clauses before presenting the meaning assembly for pied-piping inwh-questions

5331 Pied-piping in relative clauses

With a polymorphic type-assignment to wh-phrases Morrill (1994) accountsfor the pied-piping of additional material along with the wh-phrase The typeneeds to be polymorphic because the pied-piped material may form a pre-positional phrase as well as a noun phrase The following examples illustratea pied-piping construction of a noun phrase and a prepositional phrase in arelative clause7

(512) a I like the author [whose book] John read

b I like the topic [on which] John read a book

c I read the novel [the author of which ] John liked

We illustrate the meaning assembly of pied-piping constructions by ana-lyzing the complex np construction of example 512 To indicate the differentsemantic relations in the complex np we first decompose the construction andgive a brief description of the different components

(513) (I read) the novel1 [[the author of which1 ]2 John liked t2 ]7The examples are taken from Carpenter (1997)

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 31: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 183

The indices in this example indicate the semantic relations between threedifferent parts of the relative clause

bull the relative clause [[the author of which]2 John liked t2] modifies thenoun lsquonovelrsquo

bull the relative pronoun lsquowhichrsquo occupies the gap position in the direct object[the author of t1] to which the noun phrase lsquothe novelrsquo is associated

bull the gap t2 in the relative clause relates to the whole direct object NP [theauthor of which ]2

The interaction between the three parts is reflected in the syntactic typethat Morrill (1994) proposes for the analysis of pied-piping constructions8 Weuse the q-operator (Moortgat 1991) to incorporate the three parts involvedin the construction of a relative clause The meaning assembly reveals therelations between the different components of the relative clause the relat-ive clause (P) whose argument position is filled by the clause modifying thenoun phrase the relative clause (N) the common noun (V) and the gap hypo-thesis (x) which associates the common noun (V x) to the gap in the relativeclause (N x) The syntactic type of the relative pronoun along with its lexicalsemantics can be represented as follows

which q(npnp (nn)(sdiams2np))λNλPλVλx((V x)and (P (N x)))

The meaning assembly of the relative clause follows the syntactic deri-vation steps For a detailed presentation of the syntactic analysis of relativeclause constructions we refer to Morrill (1994) Carpenter (1997) We limit thepresentation here by showing the meaning that is computed for the wholeconstruction

novel the author of which John liked ` nλx((novel x)and ((like ιy((author y)and ((of x) y))) j))

5332 Pied-piping in wh-questions

The meaning assembly of the relative pronoun is very close to the meaning as-sembly of the wh-determiner for left-branch constructions The wh-questionthat we are going to derive is presented below The analysis to incorporate thesemantics of the pied-piped material in the meaning representation of a wh-question is similar to the analysis of pied-piping constructions in a relativeclause

(514) [the author of which novel1 ]2 does John like t28Morrill (1994) uses a different mechanism for deriving scope than we have discussed through-

out this thesis

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 32: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

184 5 Syntax-semantics interface of wh-questions

The indices in this example indicates the semantic relations within thecomplex np are the same as the semantic relations in the relative clause con-structions Let us now step-by step decompose the syntactic use of the wh-phrase to determine the semantic representation of pied-piping constructionsin wh-questions The different syntactic parts of the derivation are mappedto a semantic representation To formulate a lexical semantics for the pied-piping case of the which-phrases we look at the distinct components and seehow each component contributes to the meaning assembly of the whole con-struction

which novel the wh-phrase lsquowhich novelrsquo embedded inside the constructionis the actual argument phrase that is being questioned over Isolatingthe meaning assembly of the wh-phrase from the other components thesemantic representation of the wh-phrase becomes

(λVetλx(x = ιz(V z)) λy(novel y))lowast

β λx(x = ιz(novel z))

the author of which novel the np with the wh-phrase embedded in it re-stricts the semantic denotation of the wh-phrase Similar to other wh-in-situ phrases the wh-phrase is semantically represented as a lambdaterm that binds a term variable in the question body For lsquowhichrsquo thequestion body is the complex np The meaning assembly of the complexNP construction with the embedded wh-phrase can therefore be repres-ented as

(λNetλx(N x) λzιy[(author y)and ((of z) y)])lowast

β λxιy[(author y)and ((of x) y)]

wh does John like The meaning assembly of the question body which thecomplex wh-phrase will be merged with is similar to other direct ques-tions The argument variable of the predicate lsquolikersquo is bound by the λoperator

(λPetλx(P x) λu((like u) j))lowast

β λx((like x) j)

Combining the different components in the order that matches the wh-typeschema we obtain the following semantic representation for lsquowhichrsquo

which WHin(npnpWHrex(diams2np q snp))n

λVλNλPλx(x = ιz[(V z)and (P (N z))]

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 33: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 185

Sample derivation We present the analysis of lsquoThe author of which novel doesJohn likersquo in three parts the same parts as we used to explain the meaningassembly of the wh-determiner Firstly we merge the wh-determiner with thenoun lsquonovelrsquo Secondly we present part of the derivation where the wh-phrasecombines with its pied-piping material Thirdly we merge the result type ofthe complex wh-phrase to the body of the question Step-by-step we resolvethe meaning assembly of the wh-question along with the syntactic derivation

which novel

whichWHin(npnpWHr

ex(diams2np q snp))nnovel

nwhich novel ` WHin(npnpWHr

ex(diams2np q snp))[E]

λNλPλx(x = ιz[(novel z)and (P (N z) y)])

the author of which novel

which novelWHin(npnpWHr

ex(diams2np q snp))

thenpn

authorn

of(n2n)np [np]of np ` n2n

[E]

author (of np) ` 2n[E]

〈author (of np)〉 ` n[2E]

the 〈author (of np)〉 ` np[E]

the 〈author (of (which novel))〉 ` WHrex(diams2np q snp)

[WHin]

λP(λx(x = ιz[(novel z)and (P ιy[(author y)and ((of z) y)])])

wh does John like

the author of which novelWHr

ex(diams2np q snp)

doesq(np bull inf )

johnnp

likein fnp [diams2np]like diams2np ` inf

[E]

john (like diams2np) ` np bull inf[bullI]

does (john (like diams2np)) ` q[E]

(the author of which novel) (does (john like)) ` snp[WHr

ex]

λx(x = ιz[(novel z)and ((like ιy[(author y)and ((of z) y)]) j)]

The analysis of lsquothe author of which novelrsquo illustrates that we can derive acomplex wh-phrase of type WHr

ex(diams2np q snp) The meaning assembly ofthe complex np shows that we need to combine with a question body to fill in

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 34: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

186 5 Syntax-semantics interface of wh-questions

the predicate variable The last part of the derivation shows that the derivedwh-ex-situ phrase as a whole merges with the question body and replacesthe associated gap hypothesis lsquodoes John like nprsquo The meaning assembly ofthe whole sentence captures the meaning of the pied-piping construction as adirect question

5333 Piep-piping in Japanese

In chapter 4 section 425 we presented wh-questions where the wh-phraseis embedded in a complex noun phrase The syntactic analysis of these wh-questions shows many similarities to pied-piping constructions We reanalyzethe syntactic account and show that a pied-piping analysis yields a semanticsthat matches the interpretation of direct questions

Example 515 illustrates a complex noun phrase where the wh-phrase isembedded in a relative clause construction (Nishigauchi 1990 ex57) Al-though embedded in a complex NP the wh-phrase is interpreted at a mainclause level As a result the whole sentence is interpreted as a direct question

(515) Mary-waMary-[Top]

[[John-niJohn-[dat]

nani-owhat-[acc]

ageta]gave

hito-ni]man-[Dat]

atta-kamet-Q

lsquoWhat did Mary meet the man who gave to Johnrsquo

Both Pesetsky (1987) and Nishigauchi (1990) notice that a possible re-sponse to such wh-questions is to repeat the complex noun phrase where thewh-phrase is embedded in The answer given in example 516A only providesa minimal response to the embedded wh-phrase which Pesetsky (1987) in-dicates as ungrammatical Nishigauchi (1990) presents the short answer asgrammatical but claims that the short answer is a truncated form of 516BNishigauchi provides further evidence that short answers to other complexNP wh-expressions are completely unacceptable The most salient answer towh-questions where the wh-expression is embedded in a complex np mustrepeat at least some part of the complex NP Example 516B shows such apossible response

(516) A lowast Konpyuutaacomputer

desu[Cop]

lsquoItrsquos a computerrsquo

B [[Konpyuutaa-ocomputer[acc]

ageta]gave

hito]man

desu[Cop]

lsquoItrsquos the man who gave a computer (to him)rsquo

A proper analysis of Japanese complex NPrsquos must account for the semanticembedding of the wh-phrase in the relative clause In chapter 4 section 42 weshowed that complex nprsquos are correctly analyzed by assigning the embeddedinterrogative the following wh-type schema

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 35: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 187

nani WHin(npnpWHin(np q qnp))

Syntactically and semantically the type is comparable to the type assignedto wh-phrases that occur in pied-piping constructions in English The wh-phrase lsquonanirsquo (= what) occurs in an np-type construction which as a whole func-tions as a wh-in-situ For instance assume that lsquonanirsquo occurs in the complexnoun phrase expression lsquoJohn-ni nani-o ageta hitorsquo (= man who gave what toJohn) which for the sake of simplicity we abbreviate to lsquowhcomplexrsquo The com-plex wh-expression has the same syntactic type and semantic representationas other wh-phrases in Japanese

whcomplex λPλx(P x) WHin(np q qnp)

The semantic interpretation of the complex noun-phrase needs to be in-corporated into the semantic representation of the embedded wh-phrase Thesemantic representation of the wh-in-situ type becomes

nani λNλPλx(P (N x)) WHin(npnpWHin(np q qnp))

We illustrate the meaning assembly of wh-questions with a complex npFirst we derive the semantic representation of the complex np with the em-bedded wh-phrase Secondly the whole complex np is merged with the mainclause and replaces the assiociated argument position which is reflected in themeaning assembly

Complex np The syntactic analysis of complex noun phrases in Japanese isin many respects equal to the analysis of English relative clause constructionsAs we noted in chapter 4 Japanese lacks an overt complementizer such aslsquothatrsquo We therefore proposed the following type for noun phrases that arerestricted by a relative clause

hito λQetιy[(man y)and (Q y)] (NOMs)np

The derivation of the complex noun phrase lsquoJohn-ni nani-o ageta hitorsquo (= manwho gave what to John) with lsquonanirsquo as embedded wh-phrase is the following

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 36: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

188 5 Syntax-semantics interface of wh-questions

nani[[λNλPλx(P (N x))]]

WHin(npnpWHin(np q qnp))

[np]

(john ni) ((np o) ageta)[[λx(((give z) j) x)]]

NOMs

hito[[λQιy[(man y)and (Q y)]]]

(NOMs)np((john ni) ((np o) ageta)) hito ` np

[E]

[[ιy[(man y)and (((give z) j) y)]]]

((john ni) ((nani o) ageta)) hito ` WHin(np q qnp)[WHin]

λPλx(P ιy[(man y)and (((give x) john) y)])

The complex noun phrase yields the same type as a wh-in-situ phrase Themeaning assembly corresponds to the meaning assembly of a normal wh-in-situ Only the variable that refers to the gap hypothesis is embedded in thepredicate that is part of the complex np The next step is that the complex npas a wh-in-situ is merged with the body which will provide the predicate Pthat is requested by the wh-in-situ type

We use the following abbreviation to refer to the complex noun phraselsquojohn-ni nani-o ageta hitorsquo (= man who gave what to John)

whcomplex ` λPλx(P ιy[(man y)and (((give x) j) y)]) WHin(np q qnp)The following derivation illustrates the steps of merging the complex noun

phrase to the main clause After the question marker is merged with the struc-ture and forms the body of the question of type q the complex noun phrase ismerged

whcomplex

WHin(np q qnp)

mary wa

top

[np]ni

npdat(np ni) ` dat

[E] atta

dat(tops)

(np ni) atta ` tops[E]

(mary wa) ((np ni) atta) ` s[E] ka

sq(((mary wa) (np ni) atta)) ka) ` q

[E]

((mary wa) ((whin ni) atta)) ka ` wh[WHin]

λx((meet ιy[(man y)and (((give x) y) j)]) m)

The meaning assembly of the wh-question shows that the argument insidethe complex noun phrase is questioned at the main clause level The lambdaabstraction (λx) abstracts over the argument position embedded inside thecomplex noun phrase The meaning assembly of the relative clause construc-tion modifying lsquohitorsquo restricts the interpretation of the wh-phrase Possible an-swers to such a wh-questions therefore have to match the restrictive meaningrepresentation of the argument

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 37: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 189

534 Scope marking constructionsIn chapter 4 we showed how scope marking constructions in German andHindi can be analyzed syntactically using the wh-type schema The decom-position of the wh-type as a sentence which is incomplete for an answeralso affects the type assigned to the scope marker With the decompositionthe meaning assembly computed for scope marker constructions is the samemeaning assembly for long-distance wh-questions

In section 5341 we repeat the wh-type schema for German and showhow the type unfolds with the decomposition of the wh-type Step-by-stepwe build up a meaning assembly which equals that given for wh-questionswithout a scope marker In section 5342 we follow a similar line of reas-oning inspecting scope marking constructions in Hindi In section 5343 wediscuss the analyses of scope markers of Hindi and German in type-logicalgrammar in comparison to analyses for scope marking constructions that havebeen suggested in generative syntactic research (Dayal 2000 van Riemsdijk1983 McDaniel 1989)

5341 German scope marking

Scope marking constructions in German are sentences with an embedded in-terrogative clause and a left peripheral scope marker The matrix verb is abridge verb such as lsquoglaubenrsquo (= believe) that normally does not select for an em-bedded interrogative The scope marker lsquowasrsquo (= what) appears clause-inital atthe main clause while the actual wh-phrase appears clause-initial in the em-bedded clause The overall interpretation of a scope marking construction issimilar to wh-questions where the wh-phrase appears fronted The followingexamples illustrate a direct question and the same question as a scope markingconstruction

(517) Welcheswith

Bildiwhom

glaubtbelieve

MiroMiro

dassthat

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture do Miro believe that Picasso had paintedrdquo

(518) Wasiwhat

glaubtbelieves

MiroMiro

welcheswhich

Bildipicture

PicassoPicasso

ti gemaltpainted

hattehad

ldquoWhich picture does Miro believe that Picasso had paintedrdquo

In chapter 4 we assigned lsquowasrsquo the following wh-type schema

was WHlex(diams2(sprimewhprime) swh) (= wh(diams2(sprimewhprime)s))

On the basis of this wh-type schema the scope marking construction isderived as follows We only display the last steps in the derivation wherelsquowbpghrsquo is an abbreviation for an embedded interrogative clause lsquowelches BildPicasso gemalt hattersquo (= whprime)

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 38: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

190 5 Syntax-semantics interface of wh-questions

(glaubt Miro) (diams2(sprimewhprime) wbpgh) ` sdiams2(sprimewhprime) ((glaubt Miro) wbpgh) ` s

[Pl2]

[WHlex] was ` WHl

ex(diams2(sprimewhprime) swh)was ((glaubt Miro) wbpgh) ` wh

The derivation of the scope marker construction can be paraphrased inprose as follows After the embedded interrogative is built up as usual thegap hypothesis of the scope marker (diams2(sprimewhprime)) merges with the interrog-ative clause at the position where normally a complementizer such as lsquodassrsquowould be expected The scope marker functions as a lsquoleverrsquo and changes thecategory of the embedded interrogative (whprime) into the category for embed-ded declarative clauses (sprime) which in turn can be selected by the bridge verblsquoglaubenrsquo After the matrix clause is merged and the gap hypothesis is dis-placed to the left edge of the structure (Pl2) the scope marker merges withthe question body replacing its hypothesis The whole sentence becomes oftype wh

For the syntactic analysis we used wh-types in an abbreviated form Forthe semantic analysis we will unfold the wh-type inside the wh-type schemascope marker to the question-answer type snp for direct questions andsprimenp for embedded questions After unfolding we are left with the follow-ing wh-type schema for the scope marker lsquowasrsquo

wh(diams2(sprimewhprime))s)(snp)(diams2(sprime(sprimenp))s)

[un f old]

Wh-phrases and scope markers So far we have concentrated on the sim-ilarity between scope markers and complementizers However we believeand argue that the wh-scope marker is more similar to the object wh-phraseThe difference is that the object wh-phrase associates with np gap hypothesesand the scope marker associates with a sprime(sprimenp) gap Mapping the syntactictype to a semantic type reveals that the object wh-phrase reasons over gaphypothesis of type e while the scope marker is reasoning over ldquoliftedrdquo types(e rarr t) rarr t

object wh-phrase lsquowasrsquo WHlex(diams2np s snp)

scope marker lsquowasrsquo WHlex(diams2(sprime(sprimenp)) s snp)

Also semantically the similarity between object wh-phrases and scopemarkers is visible Object wh-phrases and scope markers have the followinglexical semantics Because the scope marker is reasoning over a lifted e typeinstead of applying a predicate P to the argument variable x the predicate Pis applied to the lifted argument variable λQ(Q x)

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 39: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 191

object wh-phrase lsquowasrsquo λQetλxe(Q x)scope marker lsquowasrsquo λP((et)t)tλxe(P λQet(Q x))

We illustrate the meaning assembly of scope marker constructions witha step-by step analysis and show that meaning assembly is similar to long-distance wh-questions For the analysis of scope marking constructions wefocus on the meaning assembly at the point where the scope marker is mergedwith the question body We repeat the meaning assembly for the inference rulefor merging a wh-type schema to a question body ω is the term variable thatis instantiated by the semantic term assigned to the wh-phrase

Γ ` ω WH(A BC) ∆[x A] ` BODY B∆[Γ] ` (ω λxBODY) C

[WHE]

For the derivation of the scope marking construction lsquowas glaubt Mirowelches Bild Picasso gemalt hattersquo we used the following lexical entries for thesyntactic derivation of these constructions in chapter 4 section 44 Notice thatto avoid the complexity of a verb movement analysis we have chosen to cap-ture the different orderings of arguments in main and embedded sentencesby fixing the argument ordering in the types assigned to the verb The lexicalsemantics of each word is given along with the syntactic type-assignment

glaubt λPλx((believe P) x) IVsprime

hatte λQQ prtAUXgemalt λy(paint y) npprtBild picture nPicasso Miro pm nomwelches λQλPλy((Q y)and (P y)) WHl

ex(diams2np swhprime)n

To illustrate the meaning assembly of a scope marker construction weagain build up the semantic term in a sepwise fashion For the sake of simpli-city we abbreviate the type for embedded wh-questions back to whprime (= sprimenp)

1 The embedded wh-question is built up standardly

λPλy((picture y)and (P y))

welches bild

WHlex(diams2np swhprime)

(paint x) p

diams2np (picasso (gemalt hatte)) ` s(welches bild) (picasso (gemalt hatte)) ` whprime

[WHlex]

λy((picture y)and ((paint y) p))

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 40: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

192 5 Syntax-semantics interface of wh-questions

In the following derivation we abbreviate the semantics for the embed-ded interrogative clause to λy(ppp y) and the structural occurrence oflsquowelches Bild Picasso gemalt hattersquo to wbpgh

2 The gap hypothesis of the scope marker is applied to the type of theembedded wh-question and turns the embedded wh-question type intoan embedded sentence type Semantically the scope marker hypothesisis assigned a predicate variable (Rett) which applies to the embeddedinterrogative clause yielding a proposition of type t

R[diams2(sprimewhprime) ` sprimewhprime]

λy(ppp y)wbpgh ` whprime

diams2(sprimewhprime) wbpgh ` sprime[E]

(R λy(ppp y))

3 The matrix verb selects the embedded sentence with the hypothesizedscope marker and yields a main clause structure with the scope markerhypothesis embedded

λP((believe P) m)glaubt miro ` ssprime

(R λy(ppp y))

diams2(sprimewhprime) wbpgh ` sprime

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` s[E]

((believe (R λy(ppp y))) m)

4 We have now reached the point where we can merge the scope markerto the question body The meaning assembly of the wh-type merge rulecreates a λ-binder for redrawing the gap hypothesis Additionally thelexical semantics assigned to the scope marker applies to the semanticterm of the question body The result after applying β-reduction to thesemantic term is a semantic term where the gap variable inside the em-bedded clause is bound to λ-operator which takes scope over the mainclause

λPλx(P λQ(Q x)))was

WHlex(diams2sprimewhprime s snp)

((believe (R λy(ppp y))) m)

(glaubt miro) (diams2(sprimewhprime) wbpgmh) ` swas ((glaubt miro) wbpgh) ` snp

[WHlex]

(λPλx(P λQ(Q x)) λR((believe (R λy(ppp y))) m))β λx(λR((believe (R λy(ppp y))) m) λQ(Q x))

β λx((believe (λQ(Q x) λy(ppp y))) m)β λx((believe (λy(ppp y) x)) m)

β λx((believe (ppp x)) m)

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 41: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 193

The semantics of the whole scope marking construction indicates that thescope marker associates with the gap hypothesis in the embedded interrog-ative Due to the occurrence of another wh-phrase the scope marker cannotassociate with the hypothesized argument gap directly With our proposedtype for wh-scope markers the link between the gap hypothesis and the scopemarker can be established The scope marker abstracts over the embeddedinterrogative and associates with the argument via the answer type that isrequested by the embedded interrogative therby interpreting the whole sen-tence as a direct question

Multiple scope marker construction In scope marking constructions wherethe wh-phrase occurs embedded in a clause intervened by other embeddedclauses each clause that intervenes between the embedded interrogative andthe main clause must contain another scope marker These scope markerspass the semantic representation of the embedded argument position on tothe main clause An example of a multiple scope marker construction is thesentence lsquoWas glaubte Miro was Hans meint welches Bild Picasso gemalt hattersquo (=Which picture does Miro believe that Hans claims that Picasso has painted)

In chapter 4 we showed that such constructions are derived by recursivelybinding the embedded question We derived the construction adding the fol-lowing type-assignment to scope markers that occur in the embedded clausesThe type alternates with the type assigned to the main clause scope markerAfter merging with the embedded clause the embedded scope marker returnsan embedded question type The semantic term assigned to embedded scopemarkers is the same as the semantic term for matrix clause scope markers

was λPλx(P λQ(Q x))) WHlex(diams2(sprime(sprimenp)) s sprimenp)

The derivation of a subordinate clause with an embedded scope marker issimilar to the derivation of the main clause with a scope marker Semantic-ally the scope markers in the intervening clauses pass the lambda abstractionover the term variable of the embedded interrogative on to the main clauseThe following meaning assembly is computed for deriving a multiple scope-marker construction (see appendix B55 for an on-line derivation)

(519) Was glaubt Miro was Hans meint welches Bild Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

The alternative construction where the embedded wh-phrase moves oneembedded clause higher up has the same meaning assembly In those casesthe scope marker only appears in the clauses preceding the partially movedwh-phrase The subordinate clauses that follow are headed by the comple-mentizer lsquodassrsquo

(520) Was glaubt Miro welches Bild Hans meint dass Picasso gemalt hatte` λy((believe ((claim (ppp y)) hans)) miro) snp

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 42: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

194 5 Syntax-semantics interface of wh-questions

5342 Hindi scope marking

The meaning assembly that accounts for the scope marker construction in Ger-man is similarly applicable to Hindi The difference between German andHindi is the structural realization of the scope marker In Hindi the scopemarker occurs cliticized to the matrix verb as illustrated in the following ex-ample

Scope marking construction (Dayal 2000 ex5 p160)

(521) JaunJohn

kyaawhat

soctaa haithink-[pres]

kithat

meriiMary

kis-sewho-with

baattalk

karegiido-[Fut]

ldquoWho does John think Mary will talk tordquo

lsquokyaarsquo as scope marker We proposed a similar analysis for Hindi scopemarking as for German scope marking We will show that the proposed ana-lysis for Hindi yields the same semantics for scope marking constructions asfor German

Similar to German the scope marker in Hindi is syntactically similar tothe type assigned to wh-phrase In Hindi wh-phrases may occur fronted andcliticized to the verb Because the scope marker only occurs cliticized to theverb we compare the type for scope markers with the type assigned to wh-phrase that occurs cliticized Along with the syntactic types we present lexicalsemantics on the basis of which we compute the meaning assembly of scopemarking constructions in Hindi

scope marker wh-phrasekyaa WHl

ex(diams2(sprime(sprimenp)) IVs IVwh) WHlex(diams2np IVs IVwh)

λPλxλy((P λQ(Q y)) x) λPλxλy((P y) x)where IVs = NOMs and IVwh = NOMwh

The scope marker modifies the type of the embedded interrogative andchanges the type of the embedded interrogative such that the matrix verbcombines with it Semantically it binds the embedded wh-phrase such thatit is interpreted outside the embedded clause at the main clause level Themeaning assembly of the scope marker reflects the intricate grammatical func-tion that the scope marker has

As an illustration we present an analysis of the scope marking construc-tion where the embedded wh-phrase stays in-situ We use the following lex-ical entries along with lexical semantics assigned to the types to derive a scopemarking construction In order to improve readability of the derivation weabbreviate snp and sprimenp to wh and whprime The type abbreviations IVs andIVwh stand for NOMs and NOMwh respectively

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 43: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 195

ki λQQ AprimeAs (where A isin swh)soctaahai λPλx((think P) x) IVssprime

dekhaa λyλx((see y) x) ACCIVraviine anu ra NOM

kisko λPλx(P x) WHlex(diams2ACC IVs IVwh)

We present the syntactic analysis in two steps First the derivation of theembedded questions Secondly the merging of the scope marker Finally wepresent the meaning assembly of the whole derivation

1 The embedded wh-phrase lsquoki raviine kisko dekhaahairsquo is built up as fol-lows

ki

whprimewh

raviinenom

kisko

WHlex(diams2ACC IVs IVwh)

[diams2ACC]ACC

dekhaa

ACC(NOMs)

diams2ACC dekhaa ` NOMs[E]

kisko dekhaa ` NOMwh[WHl

ex]

(raviine (kisko dekhaa)) ` s[E]

ki (raviine (kisko dekhaa)) ` whprime (= sprimenp)[E]

λx((see x) r)

2 The main clause with the scope marker lsquoAnu kyaa soctaahairsquo abstractsover the embedded wh-phrase (whprime= lsquoki raviine kisko dekhaahairsquo) as fol-lows

anunom

soctaahai

IVssprime

[diams2(sprimewhprime)]sprimewhprime

whprime

whprime

diams2(sprimewhprime) whprime` sprime[E]

soctaahai (diams2(sprimewhprime) whprime) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)kyaa (soctaahai whprime) ` NOMwh

anu (kyaa (soctaahai whprime)) ` wh (= snp)[E]

λy((think ((see y) r)) a)

3 The semantic representation of the scope marking construction is builtup as follows

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 44: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

196 5 Syntax-semantics interface of wh-questions

anua

kyaaλPλxλy((P λQ(Q y)) x)

soctaahaithink

[diams2(sprimewhprime)]κ

whprime

λz((see z) r)diams2(sprimewhprime) whprime

(κ λz((see z) r))

[E]

soctaahai (diams2(sprimewhprime) whprime)diams2(sprimewhprime) (soctaahai whprime)

[Pl2]

(think (κ λz((see z) r)))

[E]

kyaa (soctaahai whprime)λxλy((λκ(think (κ λz((see z) r))) λQ(Q y)) x)

λxλy((think (λQ(Q y)) λz((see z) r)) x)λxλy((think (λz((see z) y)) r) x)

λxλy((think ((see y) r)) x)

[WHlex]

anu (kyaa (soctaahai whprime))[E]

λy((think ((see y) j)) a)

Multiple scope markers In the same way as in German constructions withmultiple scope markers are derived by recursively binding the embeddedquestion Each embedded clause that intervenes between the embedded in-terrogative and the main clause must contain another scope marker

The syntactic type of the scope marker in the embedded clause varies fromthe main clause scope marker only in the clause type the scope marker ap-pears in sprime Overall the syntactic and semantic representation of the two scopemarkers are the same Both scope markers appear preverbally and both selectan embedded interrogative while yielding a wh-question type

intervening lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

main clause lsquokyaarsquo λPλxλy((P λQ(Q y)) x) WHlex(diams2(sprimewhprime) IVs IVwh)

Using the above type for lsquokyaarsquo in intervening clauses and the other type forlsquokyaarsquo in main clauses we derive multiple scope marking constructions Be-low we give an example of such a construction

(522) Jaun kyaa soctaahai anu kyaa kahegii ki raviine kisko dekhaa ` wh= lsquoWhom does John think that Anu says that Ravi sawrsquo

Each subpart of the clause is presented separately The two scope markersrecursively bind the argument variable of the embedded interrogative up tothe main clause

1 An embedded interrogative is selected by the hypothesized modifiertype of the scope marker lsquokyaarsquo The type modifer changes the type of

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 45: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

53 Linguistic application 197

the embedded clause from an embedded wh-question type (whprime) into anembedded declarative clause type (sprime)

[diams2(sprimewhprime)]sprimewhprime ki (raviine (kisko dekhaa)) ` whprime

diams2(sprimewhprime) (ki (raviine (kisko dekhaa))) ` sprime[E]

(κ λy((see y) r))

2 The embedded clause of the previous derivation is selected by the verbin the intervening clause Then the scope marker is inserted and with-draws the hypothesized modifier Semantically the embedded argumentphrase is now bound by the scope marker We abbreviate the embeddedinterrogative lsquoki raviine kisko dekhaarsquo to whemb

anu ` NOM

kahegii ` IVssprime diams2(sprimewhprime) whemb ` sprime

kahegii (diams2(sprimewhprime) whemb) ` IVprimes

[E][WHl

ex] kyaa ` WHlex(diams2(sprimewhprime) IVprime

s IVprimewh)

kyaa (kahegii whemb) ` IVwhprime

anu (kyaa (kahegii whemb)) ` whprime[E]

λy((say ((see y) r)) a)

3 The intervening phrase with lsquokyaarsquo in composition with the embeddedinterrogative forms as a whole an embedded interrogative The otherscope marker projects another sentence modifier which changes the typeto a embedded declarative clause (sprime)

[diams2(sprimewhprime)]sprimewhprime anu (kyaa (kahegii whemb)) ` whprime

diams2(sprimewhprime) (anu (kyaa (kahegii whemb))) ` sprime[E]

(κ λy((say ((see y) r)) a))

4 The embedded sentence is selected by the matrix verb lsquosoctaahairsquo andsubsequently the scope marker withdraws the hypothesis The wholesentence is typed as a wh-question and is interpreted as such Onceagain lsquokyaarsquo binds the lambda abstraction of the embedded itnterrogat-ive lsquoanu kyaa kahegii whembrsquo

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 46: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

198 5 Syntax-semantics interface of wh-questions

jaun ` NOM

soctaahai ` IVssprime diams2(sprimewhprime) whemb ` sprime

soctaahai (diams2(sprimewhprime) whemb) ` IVs[E]

[WHlex] kyaa ` WHl

ex(diams2(sprimewhprime) IVs IVwh)

kyaa (soctaahai whemb) ` IVwh

jaun (kyaa (soctaahai whemb)) ` wh[E]

λy((think ((say ((see y) r)) a)) j)

The proof term of the whole construction reveals that the argument whichis questioned over by the embedded wh-phrase is bound in the main clauseThis is the result of recursively applying and abstracting over the embeddedinterrogatives lsquoKyaarsquo like lsquowasrsquo in German serves as a binder operator whichreanalyzes the lambda abstractor of the clause where the scope marker is com-bined with

5343 Comparison between German and Hindi

In the generative syntactic framework different interpretational mechanismshave been proposed for the analysis of scope marking constructions The twomain approaches are the direct dependency approach (van Riemsdijk 1983McDaniel 1989) and the indirect approach (Dayal 1996 2000) The crucialdifference between these two approaches is the role of the scope marker forthe interpretation of the scope marking construction as a direct question Inthe direct dependency approach the scope marker is semantically opaque andis used only to assign matrix scope to the embedded wh-phrases In the in-direct dependency approach the embedded wh-phrases are not interpretedat main clause level but play an indirect role The embedded interrogativeforms a semantic restriction to the interpretation of the scope marker which isinterpreted as a normal wh-phrase Both approaches base their analysis on asemantic reconstruction of the phenomenon

We have argued that the semantics of wh-questions is determined by aninterface between syntax and semantics We have shown that the syntacticderivations of the two scope marking constructions is the same In both Hindiand German the scope marker associates with a gap position between the em-bedded interrogative clause and the main clause The gap hypothesis mergeswith the embedded interrogative clause and returns a declarative clause types which fulfills the selectional requirements of the matrix verb ie a bridgeverb that only selects embedded declarative clauses By identifying the scopemarker as a wh-phrase we account for the structural position of the scopemarker Additionally due to the scope markerrsquos semantic properties as a wh-phrase the meaning assembly of scope marking constructions is equal to themeaning assembly of direct questions

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 47: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

54 Concluding remarks 199

54 Concluding remarks

In this chapter we have discussed the syntactic and semantic consequencesof a structured meaning approach to wh-questions In a structured meaningapproach wh-questions are taken to be incomplete sentences that are part of aquestion-answer sequence We have proposed to decompose the wh type intoa type AB where A is the type of the question-answer sequence and B isthe type of the answer Along with the syntactic decomposition of wh-typeswe have been able to express the semantic decomposition of the semantic ω-operator as a λ-term

Additionally by incorporating the types of possible answers into the typeassigned to wh-questions we have shown that wh-type schema can be instan-tiated in a polymorphic way Based on characteristic theorems in semantictype language we have presented a derivability pattern of wh-type schemaAlong with the syntactic derivability the meaning assembly for each instanceof a wh-type schema could be computed

With the derivability pattern of wh-type schemata and the meaning as-sembly of wh-questions we have reanalyzed some phenomena of chapter3 and 4 In the analyses we have concentrated on the relation of the poly-morphic syntactic use of the wh-type schema with the meaning assembly ofwh-questions We have illustrated this polymorphic use of wh-phrases byproviding the syntactic analysis along with the derivational semantic of dif-ferent constructions multiple wh-questions long-distance binding and em-bedded wh-questions in English complex NP constructions in Japanese andscope marking constructions in German and Hindi

More specifically we have presented a uniform account of multiple wh-questions in wh-ex-situ and wh-in-situ languages The semantic differencebetween wh-pronouns and wh-determiners can be accounted for on the basisof the derivability pattern of wh-type schema For multiple wh-questions inwh-ex-situ languages wh-in-situ languages or languages such as English withsimple wh-fronting we have shown that the same meaning assembly is com-puted using structurally variant wh-type schemata Additionally we have il-lustrated that on the basis of the decomposition of wh-question types wh intotype snp the meaning assembly of scope marking constructions is equal tothe semantics of direct questions

Page 48: Syntax-semantics interface of wh-questions · As the sentences illustrate, the answers have a direct relation to the interrog-ative phrase in the question. To capture the relation

Recommended