+ All Categories
Home > Documents > ADVERBIAL COMPLEMENTS AND THEIR SCOPE PROPERTIES IN …kiss/publications/On... · ADVERBIAL...

ADVERBIAL COMPLEMENTS AND THEIR SCOPE PROPERTIES IN …kiss/publications/On... · ADVERBIAL...

Date post: 11-Sep-2019
Category:
Upload: others
View: 14 times
Download: 0 times
Share this document with a friend
38
ADVERBIAL COMPLEMENTS AND THEIR SCOPE PROPERTIES IN GERMAN * Tibor Kiss Abstract. In this paper, we discuss several grammatical properties of subcategorized preposi- tional modifiers, which we call adverbial complements. Adverbial complements typically (but not necessarily) denote spatial relations, and can be selected by stative, as well as telic causa- tive and atelic process predicates. We will refute the assumption that prepositional comple- ments necessarily lack a semantic contribution. But adverbial complements are not only inter- esting from the perspective of a distinction between governed and adjoined elements. They also show conspicuous quantifier scope variation. The analysis in terms of Head-driven Phrase Structure Grammar (HPSG) with Minimal Recursion Semantics (Copestake et al. 2005) treats adverbial complements as governed elements, which nevertheless enter the se- mantic composition independently. The pertinent scope facts will be reconciled with the anal- ysis of quantifier scope in German presented in Kiss (2001). 1. Governed Prepositions are Meaningful, Too Almost every syntactic framework introduces a distinction between case-marking preposi- tions – also called ‘governed prepositions’, or ‘prepositional objects’ – and predicative prepo- sitions. Gazdar et al. (1985:132f.) – to take one example – equate case-marking prepositions with subcategorized-for prepositions. So, predicative prepositional phrases are classified as optional modifiers bearing content, while case-marking prepositions head semantically vacu- * This paper goes back to presentations given at HPSG 2011 (University of Washington, Seattle) in August 2011 and the conference Explorations in Syntactic Government and Subcategorization at the University of Cambridge in Au- gust/September 2011. I would like to thank Antje Müller, Stefan Müller, Claudia Roch, Anna Kibort, as well as two anonymous reviewers for their comments and suggestions, and finally Johanna Poppek for her help with the illustrations.
Transcript

ADVERBIAL COMPLEMENTS AND THEIR SCOPE PROPERTIES IN GERMAN*

Tibor Kiss

Abstract. In this paper, we discuss several grammatical properties of subcategorized preposi-

tional modifiers, which we call adverbial complements. Adverbial complements typically (but

not necessarily) denote spatial relations, and can be selected by stative, as well as telic causa-

tive and atelic process predicates. We will refute the assumption that prepositional comple-

ments necessarily lack a semantic contribution. But adverbial complements are not only inter-

esting from the perspective of a distinction between governed and adjoined elements. They

also show conspicuous quantifier scope variation. The analysis in terms of Head-driven

Phrase Structure Grammar (HPSG) with Minimal Recursion Semantics (Copestake et al.

2005) treats adverbial complements as governed elements, which nevertheless enter the se-

mantic composition independently. The pertinent scope facts will be reconciled with the anal-

ysis of quantifier scope in German presented in Kiss (2001).

1. Governed Prepositions are Meaningful, Too

Almost every syntactic framework introduces a distinction between case-marking preposi-

tions – also called ‘governed prepositions’, or ‘prepositional objects’ – and predicative prepo-

sitions. Gazdar et al. (1985:132f.) – to take one example – equate case-marking prepositions

with subcategorized-for prepositions. So, predicative prepositional phrases are classified as

optional modifiers bearing content, while case-marking prepositions head semantically vacu-

                                                                                                               * This paper goes back to presentations given at HPSG 2011 (University of Washington, Seattle) in August 2011 and the

conference Explorations in Syntactic Government and Subcategorization at the University of Cambridge in Au-gust/September 2011. I would like to thank Antje Müller, Stefan Müller, Claudia Roch, Anna Kibort, as well as two anonymous reviewers for their comments and suggestions, and finally Johanna Poppek for her help with the illustrations.

ous prepositional complements, whose form is determined by an external governor (as e.g.

proposed in Haider 2010:252). Similarly, Pollard and Sag (1994:255) employ the idea of an

identity function of the preposition (which makes them semantically vacuous) in their analysis

of reflexive binding into prepositional complements. We agree that a lack of semantic content

can be observed for a subset of prepositional complements. However, it should be taken as a

characteristic property of prepositional objects in general. Indeed, Bierwisch (1988), Maien-

born (1990), and Dowty (2003) have already challenged the distinction between case-marking

and content-bearing prepositions. Dowty (2003) employs the term subcategorized adjunct.

Dowty’s analysis is similar in spirit to the present one, yet we refrain from using this term.

The main reason is that we want to reserve the term adjunct to elements that are never obliga-

tory (cf. Ackema 2015:265) – which is not correct for adverbial complements.

1.1. Adverbial Complements

Examples of adverbial complements of stative, causative and process predicates are given in

(1), (2), and (3).

(1) Ein Schimmer lag [PP über dem gesamten Bild].

a gleam lay above the whole picture

‘the whole picture was gleaming’

(2) Er stellte das Buch [PP auf den Tisch].

he put the book at the table

‘he put the book on the table’

(3) Sie ziehen maschinell eine Sprengschnur [PP durch den Abschnitt].

they pull mechanically a detonating cord through the section

‘they pulled a detonating cord through the section by use of a machine’

The semantic contribution of the prepositions is a spatial one. In (1), the PP denotes a cover-

ing relation such that the picture is covered by the gleam. In (2), the PP denotes the result of a

change of position. The PP in (3) denotes a spatial traversal. The PPs are not restricted to be

headed by a particular preposition. Also, they can be replaced by other prepositions that ex-

press spatial senses. This is illustrated for the verb liegen ‘to lie’ in (4):

(4) a. Ein Tuch lag über dem Buch.

a scarf lay over the book

‘there was a scarf on top of the book’

b. Ein Tuch lag unter dem Buch.

a scarf lay under the book

‘there was a scarf below the book’

c. Ein Lesezeichen lag zwischen jeder Seite des Buchs.

a bookmark lay between every page the book

‘a bookmark could be found between each page of the book’

Although the verb does not restrict the lexical choice of the PP’s head, adverbial complements

are more or less obligatory, as is illustrated with the following examples.1

                                                                                                               1 An anonymous reviewer has noted that intransitive uses of stative verbs are not completely forbidden, as e.g. illustrated

by (i) and (ii): (i) Das Bild hängt jetzt wieder. the picture hangs now again ‘the picture is hanging on the wall again’ (ii) Das Baby steht schon. the baby stands already. ‘the baby is already able to stand by himself’ Maienborn (1990) has argued that such intransitive uses (which imply existential closure of the locative argument) are

only possible if a suitable contrast is provided either by the situation in which the sentence is uttered, or by linguistic means (as in the examples above, where we either talk about the fact that the picture is now located at the wall again, and that the toddler is already able to stand instead of just crawl around on all fours). We conclude that adverbial comple-ments of stative (and causative) predicates are obligatory unless such contextual conditions apply.

(5) a. *Ein Schimmer/Ein Tuch/ Ein Lesezeichen lag.

a gleam /a scarf/ a bookmark lay

b. *Sie stellen das Buch.

they put the book

c. Sie ziehen maschinell eine Sprengschnur.

they pull mechanically a detonation cord

‘they pulled a detonating cord by use of a machine’

Obligatoriness may be considered indicative for a complement relation, but unfortunately

does not serve as a clear criterion to distinguish complements from modifiers (cf. Christie

2013 for a recent assessment).2 We note that the PP is obligatory in (5a, b), but not in (5c),

and yet we aim to analyse all three cases by the same means.3 Similar considerations apply to

semantic completeness, i.e. the idea that an optional element has to be considered as argument

if it can be inferred from the predicate when absent. With regard to the predicates in (1), (2),

and (3), we can observe that a stative localization must position an element into one or the

other area, and a change of position leads to the inference that the object that changed position

will occupy a new one. Finally, if something is pulled, it will follow a path, even if the path is

not explicitly mentioned. As it stands, the inferences drawn here are not general ones, but

ones that are specific to the predicates involved, and hence, we see a justification to speak of

complements even in the case of (5c). To take up Ackema’s (2015:265) quote again, comple-

ments may be optional, but adjuncts should never be obligatory.

Before we will discuss how adverbial complements differ from governed PPs, we would

briefly like to sketch the present analysis. Adverbial complements are syntactically selected

                                                                                                               2 One must in fact ask whether obligatoriness is related to the complement/modifier distinction at all if cases outside the

verbal realm are considered, as e.g. the apparent obligatoriness of determiners with singular count nouns. 3 An anonymous reviewer has suggested treating adverbial complements of process predicates as resultative phrases.

Clearly, this assumption would account for the optionality of the phrase (cf. Müller 2002). Yet, we refrain from applying such an analysis for the following reasons: First, the PP modifying the process predicate does not describe the result of the action denoted by the verb, but a path which will be followed while the action is carried out (so being in the path will not necessarily be a result of the action). Secondly, it would be necessary to account for the observation that this particu-lar resultative construction can only be built with PPs and not with APs. Finally, even if an analysis of the process predi-cates as resultative would be pursued, it would not account for the behaviour of stative and causative predicates, which require the presence of the PP in any case.

by verbal predicates. The selection includes the syntactic category, and a restriction on the

semantics of the PP. So stative predicates and telic causatives select (stative) localizations,

process predicates select paths. The governing predicate does not impose further requirements.

The important difference to a simple complement analysis of the PPs comes into play when

the semantic contribution of the governing predicates is considered: We do not assume that

liegen (reside) introduces a binary relation, and neither do we assume that ziehen (pull) or

stellen (put) introduce ternary relations. Instead, we assume that the composition of the se-

mantics of the unary and binary predicates with the semantic contribution of the PPs takes

place in the lexicon. From this perspective, adverbial complements behave like modifiers, the

only difference being that the intersection will already be imposed in the lexicon. The present

analysis shares the insight of Dowty (2003) that a modification analysis within the lexicon

imposes a lesser burden to language acquisition, and although the present analysis differs in

terminology and in its implementation from Dowty’s analysis within Categorial Grammar, we

share the assumption that adverbial complements do not show the ordinary semantic behav-

iour of syntactic complements.

1.2. How Case-Marking Prepositions Differ from Adverbial Complements

Case-marking prepositions (governed PPs, prepositional objects) differ in various respects

from adverbial complements. Prepositional objects without semantic content can perhaps best

be illustrated by considering predicates that alternate between a prepositional and a nominal

complement without discernible changes in interpretation, as can be witnessed in (6).

(6) a. Ulrich erinnerte sich an den Vorfall.

Ulrich remembered REFL at the incident

b. Ulrich erinnerte sich des Vorfalls.

Ulrich remembered REFL the.GEN incident.GEN

‘Ulrich recalled the incident’

Since the two examples in (6) are near synonymous, it seems plausible to assume that the

preposition does not contribute to the meaning of the phrase. One can further argue that the

form of the preposition is selected in the same way as the case of the NP, and hence must be

invariant.

In some cases, a predicate may select different prepositions, leading to different inter-

pretations of the predicate. An example is the verb freuen ‘to be pleased’. This verb can be

used intransitively, but it can also be combined with the prepositions über ‘about’, an ‘on’,

and auf ‘to’.

(7) a. Er freute sich über das Spiel.

he was.pleased REFL about the game

‘he was pleased about the game’

b. Er freute sich an dem Spiel.

he was.pleased REFL on the game

‘the game gave him pleasure’

c. Er freute sich auf das Spiel.

he looked.forward REFL on the game

‘he looked forward to the game’

Example (7a) shows a sense of the preposition über ‘about’ that can be paraphrased as topic

or theme. This sense is already inherent in the interpretation of the predicate, but if we as-

sumed that the preposition actually carried that meaning, it would at least be compliant with

the sense of the predicate.

The preposition an in (7b) may be said to show a causal interpretation, but this does not

usually occur with an if an is used as a head of a predicative PP. It seems to emerge from a

combination with certain psychological predicates such as sich ergötzen ‘to amuse oneself’,

erkranken ‘to get ill’, sich (er)freuen ‘to be delighted’, sich laben ‘to refresh’, leiden ‘to suf-

fer’, sterben ‘to die’, and zerbrechen ‘to break in parts’. Since the causal interpretation can

only emerge in the context of the aforementioned predicates, it seems justified to assume that

the preposition is actually governed, and does not contribute a meaning in itself.

The preposition auf in (7c) lacks a semantic contribution, if only for the reason that

none of the usual senses of auf can be said to apply to (7c). The predicate, however, changes

its meaning if it combines with an auf-PP.

So if a lack of semantic contribution (or an interpretation that is confined to the pres-

ence of an external governor) would be a defining instance of prepositional objects, the PP in

(7c) would be analysed as a prepositional object since no discernible sense can be attributed

to the preposition, the PP in (7b) could be analysed as a governed PP if we assume that the

predicate yields the additional causal interpretation, and marks this by selecting the preposi-

tion an. The PP in (7a) would be classified as an adverbial modifier, but one that provides a

sense already to be inferred from the predicate.

Mirroring the optionality of adverbial complements, government of prepositions does

not imply obligatoriness of the complement. The verb erinnern ‘remember’ in (6) does not

require the syntactic realization of its genitive object, nor of the governed PP. In other cases,

though, governed PPs become obligatory. Illustrations are given in (8).4

(8) a. Er freute sich *(auf das Spiel).

He looked.forward REFL on the game

‘he looked forward to the game’

b. Er dachte *(an einen anderen Dozenten).

He thought on an other instructor

‘he thought of another instructor’

                                                                                                               4 It has sometimes been proposed that the an-PP in (8b) could have a causal interpretation, aligning it with the an-PP in

(7b). Zifonun et al. (1997:2154) argue against this view; I cannot detect any causality in (8b).

The presence of the auf-PP is obligatory in (8a) if the sense of the verb should be retained. If

the PP is omitted in (8a), the intended interpretation (‘to look forward to’) is no longer availa-

ble (neither can it be inferred from the predicate). Omitting the an-PP in (8b) simply leads to

ungrammaticality.

Coming back to the semantic contribution of adverbial complements and governed PPs,

we note we note that the prepositions in (1), (2), and (3) can be adverbially modified, which is

taken to be a characteristic property of predicative prepositions since Jackendoff (1973). This

is illustrated in (9).

(9) a. [PP Nahezu über dem gesamten Bild] lag ein Schimmer.

almost above the whole picture lay a gleam

‘the picture was glistening almost completely’

b. [PP Unmittelbar auf den Boden] stellten sie das Gehäuse.

directly on the floor put they the casing

‘they put the casing directly on the floor’

c. [PP Quer durch den Abschnitt] wird eine Sprengschnur gezogen.

across through the section AUX.PASS a detonating cord pulled

‘a detonating cord is pulled right across the section’

That nahezu ‘almost’, unmittelbar ‘directly’, and quer ‘across’ in (9) modify the P(P) and not

the verb, can be seen from the topicalizations in (9). If the adverbials would modify the verbs,

the constructions in (9) would violate the verb second constraint, according to which only one

constituent may precede the finite verb in a German matrix clause.

The aforementioned examples further corroborate the claim that adverbial complements

are meaningful, and that they act as complements at the same time. Any analysis of these

phrases must take this dual nature into account.

Adverbial modification, however, is impossible for governed prepositions. This can be

illustrated by considering the preposition über ‘over/above’. This preposition shows a spatial

sense of verticality, which is neutral with regard to contact, as is illustrated in Fehler! Ver-

weisquelle konnte nicht gefunden werden.), where the PP is modified by unmittelbar ‘di-

rectly’. Unmittelbar, however, cannot modify a purely synsemantic use of über, as is illustrat-

ed in (10a). The topicalization of the phrase unmittelbar über das Spiel leads to ungrammati-

cality in (10b).

(10) a. *Er freute sich unmittelbar über das Spiel.

he looked.forward refl directly about the game

b. *Unmittelbar über das Spiel freute er sich.

directly about the game looked.forward he REFL

Finally, the semantic nature of adverbial complements prohibits the realization of adverbial

modifiers with the same semantics in the same clause, as can be witnessed in contrasting (11a)

with (11b).5

(11) a. Auf der Party freute er sich auf die Verabredung.

on the party looked.forward he REFL on the date

‘at the party, he looked forward to the date’

b. *Über dem gesamten Bild lag ein Schimmer über dem Rahmen.

above the whole picture lay a gleam above the frame

1.3. Scopal Properties of Adverbial Complements

Adverbial complements of the type presented in (1) and (3) show a somewhat surprising sco-

pal behaviour, which can be illustrated by considering example (4c) above. In its most natural

reading, the adverbial complement receives wide scope over the subject in (4c), and this can-

not be accounted for by assuming that ∃∀ readings imply ∀∃ readings: it is clearly a different

bookmark that is found between each page. This observation goes against a tradition estab-

                                                                                                               5 The constraint that actually blocks the realization of more than one adverbial with the same interpretation in the same

clause is referred to quite often, but as yet, has not received a formal integration into existing frameworks.

lished since Frey (1993), and further elaborated in Kiss (2001), and Sauerland and Elbourne

(2002) among others. This line of research has shown that wide scope object quantification in

German is much more restricted than wide scope object quantification in English. While ob-

ject quantifiers in English may receive wide scope interpretations without further proviso,

object quantifiers in German require either topicalization (i.e. syntactic realization in front of

the finite verb in verb-second clauses) or scrambling (local dislocation in the so-called Mittel-

feld) to allow wide scope readings. As Kiss (2001) has proposed, a quantifier Q1 can only

outscope another quantifier Q2 if Q1 is either more prominent than Q2 in terms of syntactic

configuration, or in terms of the argument structure. Prepositional objects behave like NP

objects of transitive verbs in this respect, as is illustrated in (12) and (13).6

(12) Jeder Mann freut sich auf eine Verabredung.

every man look.forward.3SG REFL on a date

‘every man looks forward to a date’ √∀∃, *∃∀

(13) Auf eine Verabredung freut sich jeder Mann.

on a date look.forward.3.SG REFL every man

‘every man looks forward to a date’ √∀∃, √∃∀

The scopal relations in (12) are fixed: the subject has been topicalized (and is thus configura-

tionally most prominent). and it occupies a more prominent position on the argument structure

of the predicate. The situation is different in (13), where the prepositional object has been

topicalized. While it is still less prominent than the subject on the argument structure, it now

occupies a more prominent position than the subject in the syntactic configuration. Since

prominence determines scopal relations, the example is predicted to be ambiguous between a

narrow scope and a wide scope reading of the object.

                                                                                                               6 The analyses of Frey (1993) and Kiss (2001) do not take effects of intonation into account. As has been reported in

Krifka (1998), objects may indeed outscope subjects given the so-called Hat contour. In line with the present analysis, the marked intonation of the Hat contour acts as a signal to indicate scope variation.

As we have already indicated for (4c), this conclusion is not warranted for adverbial

complements: they differ from governed PPs in that wide scope object quantification becomes

possible without ostensible scrambling or topicalization of the adverbial complement. The

most natural wide scope reading of the adverbial complement could not be derived if the sub-

ject is more prominent than the adverbial complement in terms of syntactic configuration and

argument structure. These considerations do not only apply to stative predicates like liegen

‘to lie’, but also to process predicates like ziehen ‘to tow/to pull’, and causatives like stellen

‘to put’, as can be witnessed in (14) and (15).

(14) Sie ziehen eine Schnur durch jeden Abschnitt.

they pull.3PL a cord through every section

‘they pulled a cord through every section’ √∀∃, √∃∀

(15) Sie stellen jede Flasche in einen Kasten.

they put.3PL every bottle in a crate

‘they put every bottle in a crate’ √∀∃, √∃∀

The most plausible reading of (14) is one where jeden Abschnitt ‘every section’ receives wid-

er scope than eine Schnur ‘a cord’.7 Example (15) is genuinely ambiguous between the ∃∀

and the ∀∃ interpretation. Given the assumptions in Frey (1993) and Kiss (2001), the wide

scope readings of the PPs in (4c), (14), and (15) are surprising at first. Yet an analysis can be

proposed that reconciles the aforementioned analyses with the data presented here. According

to this analysis, we will assume that adverbial complements must always be more prominent

than one other argument of the predicate. For (14) and (15), this implies that the PPs are more

prominent than the accusative NPs, and that the latter have been scrambled. While this as-

sumption may be uncontroversial, the same considerations require that the adverbial comple-

                                                                                                               7 As with example (4c) one could argue that a wide scope reading of eine Schnur (a cord) actually entails a wide scope

reading of jeden Abschnitt (every section), since ∃∀ ⇒ ∀∃. But, similarly to (4c), the most plausible reading of (14) en-tails that the cord actually varies with every section, i.e. that it is not the very same cord that is pulled through every sec-tion.

ment is more prominent in terms of argument structure than the subject in (1) and (4). This

assumption clearly is controversial.

In section 2, we will discuss how to deal with wide scope readings in German in an

HPSG analysis with Minimal Recursion Semantics (MRS; Copestake et al. 2005), basically

following an analysis presented in (Kiss 2001).8 The analysis of adverbial complements in

section 3 combines the insights of the prior sections and will particularly discuss whether the

position of adverbial complements on the argument structure of their governors can be inde-

pendently justified. A brief summary closes this paper in section 4.

2. Configurational and Relational Scope Determination in German

It is a peculiar property of adverbial complements that they show quantifier scope options in

the apparent absence of overt dislocation. In (4c), the subject has been topicalized, and yet it

allows an interpretation according to which the bookmarks vary with the pages of the book. In

(14) and (15), we do not see dislocation at all, and yet the example are claimed to be ambigu-

ous. Before we can address why (4c) shows the pertinent reading, and why (14) and (15) are

ambiguous, we must review how quantifier scope is dealt with in German.

2.1. Scrambling and Quantifier Scope in German

A long tradition of quantifier scope analyses assumes that scope in English is independent

from the surface structure of the scopal elements (cf. May 1985). But not all languages follow

the same strategy with respect to scope determination. Hoji (1986) and Joo (1989) have

                                                                                                               8 The present choice for an analysis in terms of HPSG and a non-movement treatment of scrambling is not just an issue of

personal preference. As is discussed in Kiss (2001:145ff.), a movement- (and trace-) based account systematically allows derivations of more readings than are acknowledged in German.

shown for Japanese and Korean, respectively, that scope determination is constrained by the

surface position and the prominence of the quantified arguments. Japanese and Korean share

with German that arguments can be scrambled as well as topicalized, and scrambling and top-

icalization enter crucially into scope determination, as has been subsequently discussed in

Frey (1993), and Kiss (2001).

Basically, the aforementioned authors observed that arguments realized in unmarked

order in the Mittelfeld do not lead to scope ambiguity, as is illustrated for transitive structures

in (16).

(16) a. Ich glaube, dass irgendein Mann jedes Buch gelesen hat. (∃∀, *∀∃)

I believe that some man.SBJ each book.OBJ read has

‘I believe that some man has read every book’

b. Ich glaube, dass jedes Buch irgendein Mann gelesen hat. (∀∃, ∃∀)

I believe that every book.OBJ some man.SBJ read has

‘I believe that every book has been read by some man’/’I believe that some man

has read every book’

If subject and object appear in this order in the Mittelfeld, as in (16a), there is only one read-

ing available, in which the subject receives scope over the object.9 In (16b) the object has

been realized to the left of the subject. Frey (1993) assumes that this was due to movement,

and that the availability of traces introduces the ensuing scope ambiguity: Frey proposes the

Scope Principle according to which a quantifier phrase α may take scope over a quantifier

phrase β if α either c-commands β or a trace of β. The subject quantifier c-commands the

trace of the object quantifier, while the object quantifier c-commands the subject quantifier,

thus giving rise to two readings. Kiss (2001:145ff.) criticises the aforementioned proposals

since they give rise to spurious readings, and suggests replacing traces by the relative order-

                                                                                                               9 Since ∃∀ implies ∀∃, it should be stressed that the sentence does not allow an interpretation where the denotation of

Mann varies with the denotation of Buch.

ing of the quantifiers on the argument structure of their verbs. We will briefly review this

proposal and then discuss how adverbial complements fit into the picture.

The analysis of Kiss (2001) is based on HPSG (Pollard and Sag 1994) with the addition

of Minimal Recursion Semantics (Copestake et al. 2005).

The present analysis does not only rely on the semantic representation, but also on the

feature ARG-ST (argument structure). In HPSG, syntactic arguments of a predicate are found

on this list-valued feature. The feature ARG-ST thus serves as an intermediary between the

lexical semantics of a predicate and its requirements on syntactic complements. It contains the

syntactic arguments of a predicate and also determines the relative prominence of the argu-

ments. In some models, it is assumed that ARG-ST is present on lexical entries only, but the

present analysis makes use of ARG-ST to determine scope and hence ARG-ST is present on

phrases, too.

Most importantly for the present analysis, the position of an argument on ARG-ST re-

places traces in this proposal. Accordingly, Frey’s scope principle receives the following ren-

dering:

(17) Quantifier Scope in German

The scope of a quantifier is either the semantic contribution of the syntactic sister of

the quantifier or the semantic contribution of a more oblique co-dependent of the

quantifier. (Kiss 2001:157)

The condition in (17) can be illustrated with the schematic analysis of (16b) in (18). In (18),

the two quantifiers bear the subscripts 1 and 2, for subject and object, respectively, and these

two positions on the ARG-ST list of the predicate indicate that the object is a more oblique

co-dependent of the subject.

(18) Ich glaube, dass [jedes Buch irgendein Mann gelesen hat]

Before we can explain the scope variation in (18), we have to consider how scrambling is

analysed in the present framework. Following the proposal in Müller (2015), we assume that

scrambling is not dealt with through the introduction of SLASHes, but by taking elements

from any position of the COMPS (COMPLEMENTS) list (which again is filled with the ele-

ments of ARG-ST).10 Since a complement can be taken from any position of COMPS, the

label VP in (18) does not indicate the – customary – phrase resulting from a combination of a

verb and its object, but merely serves as an abbreviation for a verbal projection that is not

fully saturated. Similarly, S stands for a fully saturated verbal projection.

Let us assume that the subject takes the semantic contribution of its syntactic sister as its

scope. In this case, the scrambled object quantifier can only take the semantic contribution of

the VP as its scope. This contribution, however, is identical to the semantic head of the VP,

which is the semantic contribution of the subject quantifier. Hence, the object quantifier takes

the subject quantifier in its scope (technically, this is spelled out in section 2.2.).

Alternatively, the subject can also take the semantic contribution of the object as its

scope, since the object is a more oblique co-dependent, leading to a narrow scope reading of

the object. The example is thus predicated to be ambiguous, and the ambiguity can only

emerge if the object is dislocated.

                                                                                                               10 Both topicalization and scrambling lead to extended scope options. Yet, they must be kept apart. Topicalization belongs

to the group of unbounded dependency constructions, while scrambling is strictly local – in German at least. It is thus not surprising that the SLASH feature handles topicalization, and scrambling is dealt with by extracting elements from the COMPS list in arbitrary order.

SARG-ST<1, 2>

VPARG-ST<1, 2>

VARG-ST<1, 2>

gelesen ARG-ST<1, 2> hat

NP1

irgendein Mann

NP2

jedes Buch

1

2.2. Minimal Recursion Semantics

The basic semantic units for predicates in MRS are elementary predications (EPs). EPs con-

sist of a relation name, a label, and several argument slots. Labels serve to embed EPs under

other EPs without formal subordination (cf. below). The EP of the English quantifier every is

given in (19): the relation name is every_rel, the label of the quantifier is h0, its arguments

slots are called BV (for bound variable), RESTR, and SCOPE (for restriction and scope).

(19) every

If lexemes and phrases are combined syntactically, their EPs are combined under a set-valued

feature called RELS (for relations). Such a flat combination of EPs, however, would not be

sufficient to determine subordination and scope options. To this purpose, Copestake et al.

(2005) distinguish between ordinary and scopal argument slots. Ordinary argument slots are

linked to indices of other relations, but scopal argument slots are linked to the labels of other

EPs. Hence their value is not a variable but a handle.11 The effect of this distinction is that

intersective modification can be modelled by identifying ordinary argument slots, but subor-

dination will be modelled by identifying handles of other relations as values for scopal argu-

                                                                                                               11 A completely flat system that would only rely on coordinating elementary predications would not be able to distinguish

between the nuclear scope and the restriction of a quantifier, and similarly for other scope-bearing elements. Copestake et al. (2005) introduce handles to keep the system as flat as possible, while at the same time capturing the crucial distinction between restriction and scope.

2

6666664

every relLABEL h0

BV xRESTR h1

SCOPE h2

3

7777775

1

ment slots. In (19), BV is an ordinary argument slot and RESTR and SCOPE are scopal ar-

gument slots.

In order to keep track of the relation that has introduced a label, and to identify indices

for selection, the semantic representation in MRS makes use of the following features:

The KEY feature refers to the EP that has been introduced by the lexical head of a pro-

jection. As was already mentioned, the EPs collected during the syntactic combination are

combined in the feature RELS. Finally, the feature HOOK is necessary to identify the variable

that can currently be modified intersectively, and the handle that can be selected as a semantic

argument. The latter element is also called LTOP, standing for the local top handle.

The pertinent features to model the quantifier every are provided in (20).

(20) every

To make the semantic combination slightly more transparent, we will employ a linear notation

in the following:

• The RELS feature is provided as a flat list containing EPs: { … }.

• EPs are represented as follows: a label (hn) is prefixed to the relation name of the EP, fol-

lowed by ordinary argument slots, followed by scopal argument slots.

• The values of ordinary argument slots are indices, referring to individuals or events. Indi-

ces are represented as variables (e, x, y, z). The values for scopal argument slots are labels,

represented through hn.

2

6666666666666664

CONT

2

6666666666666664

HOOK

"LTOP h0

INDEX x

#

KEY 1

RELS

8>>>>>><

>>>>>>:

1

2

6666664

every relLABEL h0

BV xRESTR h1

SCOPE h2

3

7777775

9>>>>>>=

>>>>>>;

3

7777777777777775

3

7777777777777775

1

• The KEY is represented by underlining.

• The values of the HOOK are provided in parentheses and separated from the remaining

features by a vertical bar.

The semantic representation of the English NP every old man in (21) illustrates this.

(21) every old man

(x, h0)|{h0:every(x, h1, h2), h1:man(x), h1:old(x)}

The semantic representation in (21) starts with the HOOK, i.e. (x, h0). Then, the RELS feature

follows, containing three EPs, the second of which is the KEY. In the case of every, the rela-

tion introduces three argument slots. The first argument slot of every is an ordinary argument

slot, the second and the third argument slots are scopal argument slots that take handles as

their argument. The names of the argument slots can be introduced inside an EP, if their

names serve to make the representation more transparent:

(22) h0:every(BV: x, RESTR: h1, SCOPE: h2)

In (21), we see that the restriction of the quantifier is determined by arguments bearing the

handle h1, while the scope (h2) is not yet determined. The two other elementary predications

introduce ordinary argument slots only that identify the variable bound by the quantifier’s

elementary predication. That both elementary predications share the same handle further indi-

cates that they have been combined semantically by intersection.

The principles in (23) and (25) govern the semantic combination. Both principles make

crucial use of the HOOK features, and of the concept of a semantic head. We will begin with

the Semantics Principle in (23).

It fulfils two tasks. In its first clause, it determines the semantic head of a phrase; in its

second clause, it guarantees that EPs present in the daughter of a phrase will be present in the

mother as well.

(23) Semantics Principle:

a. The index of a phrase XP is identified with the index of the semantic head of XP.

The semantic head of XP is the adjunct daughter if the LTOP of the head daughter

is identified with one of the scopal argument slots of the adjunct daughter’s KEY

elementary predication. In all other cases, the syntactic head is the semantic head.

b. The set of elementary predications of a phrase XP is given as the set-union of the

values of each of XP’s daughter’s set of elementary predications.

We will briefly illustrate the workings of the Semantics Principle by looking at the combina-

tion of an adverbial modifier with a verb. Following Pollard and Sag (1994), intersective

modifiers show the feature MOD as part of their HEAD features. In combination with the

Head-Modifier Schema (Pollard and Sag 1994:56) this feature serves to identify the variable

of the element that is modified, and thus caters for the intersective semantics of the modifier.

Given the modifications suggested in Copestake et al. 2005, the MOD feature will select the

LTOP and INDEX features (i.e. the HOOK) of the element that is modified, and will require

that the LTOP and the INDEX be identified with its own LTOP and INDEX, respectively. In

the following structure, we will only analyse the combination of the PP im Haus (‘in the

house’) with the verbal participle geküsst (‘kissed’):

(24) Er hat sie [VP im Haus geküsst].

He have.3SG her in house.DAT kiss.PART

‘he has kissed her in the house’

VP(e, h0)|{h0:kiss(e, x, y), h0:in(e, h2),h2:house(z)}

V(e, h0)|{h0:kiss(e, x, y)}

gekusst

PP(e’, h1)|{h1:in(e’, h2), h2:house(z)}MOD:V[(e, h0)=(e’, h1)]

im Haus

1

We have indicated the requirement that the event variable and handle of the modified ele-

ments must be identified with the event variable and the handle of the modifier through the

equation (e, h0) = (e’, h1), as the semantic value of the MOD feature, while syntactically, it

selects a verbal projection, indicated through V.

The first clause of the Semantics Principle identifies the syntactic head of the phase, i.e.

the verbal participle as the semantic head as well. Hence, the phrasal projection VP bears the

LTOP (e, h0). The RELS sets of the two daughters are combined into the RELS set of the

phrase, following the requirements of the second clause of the Semantics Principle.

The Handle Projection Constraint (HPC) in (25) constrains the identification of the

LTOP in cases where scopal elements enter into the structure.

(25) Handle Projection Constraint (HPC):

If the LTOP of one of the daughters of a phrase is selected as scopal argument of an-

other daughter, the LTOP of the phrase is identified with the LTOP of the scope-

taking daughter. In all other cases, the LTOP of a phrase is identical to the LTOP of

the semantic head of the phrase.

Following Kiss (2001), we will combine the HPC in (25) with the condition on quantifier

scope in (17) to yield scope options depending on the syntactic configuration and the position

on the argument structure of the quantifier. As an illustration, we will consider the analysis of

(16a) in (26).

(26) Ich glaube, dass irgendein Mann jedes Buch gelesen hat.

The HPC identifies the LTOP of a phrase, depending on whether one of the daughters of the

phrase has selected the other daughters’ LTOP as its scopal argument. The combination of the

object quantifier with the predicate is a case at hand, just as the combination of the subject

quantifier with the VP.

The object quantifier takes the predicate as its scopal argument, and hence the respec-

tive labels are identified. (h3 = h5). The handle of the subject quantifier becomes the LTOP

after combination with the VP (h2 = h7). The index of the predicate remains the active index

throughout, since the predicate is the semantic (and syntactic) head of the clause. The second

clause of the Semantics Principle guarantees that all EPs are combined.

Considering the argument structure of the predicate, (26) is predicted to receive a

unique interpretation where the subject takes scope over the object. The analysis of (16b)

should yield a different result, as indicated in the analysis in (27).

S(e, h

1

)|{h

1

:some(x, h

6

, h

2

),

h

6

:man(x), h

2

:every(y. h

4

, h

5

),

h

4

:book(y), h

5

:read(e, x, y)}

VP(e, h

2

)|{h

2

:every(y, h

4

, h

5

),

h

4

:book(y), h

5

:read(e, x, y)}

V(e, h

3

)|{h

3

:read(e, x, y)}

gelesen

(e, h

3

)

hat

NP(y, h

2

)|{h

2

:every(y, h

4

, h

5

),

h

4

:book(y)}

jedes Buch

NP(x, h

1

)|{h

1

:some(x, h

6

, h

7

),

h

6

:man(x)}

irgendein Mann

1

(27) Ich glaube, dass [Q2 jedes Buch] [Q1 irgendein Mann] gelesen[ARG-ST <1, 2>] hat.

The analysis is (27) currently leaves open how the SCOPE values of the quantifiers – h5 and

h7 – are determined. But now, there are two options: when the subject quantifier is combined

with the predicate, it can either take the label of the predicate (h3) as its SCOPE value (h3 =

h7), or the label of its more oblique co-dependent (h2 = h7). Let us assume that the second op-

tion applies. In this case, the HPC predicts that the handle of the VP becomes the LTOP of the

phrase, and consequently that this handle can be selected by the object quantifier (h5 = h3).

Hence, the scopal relations are resolved as follows: the subject takes scope over the object

(h7 = h2), and the object takes scope over the predicate (h5 = h3).

Let us then assume that the subject takes the handle of the VP as its scopal argument (h7

= h3). The HPC requires that the handle of the subject (h1) become the LTOP of the VP, and

the object is taking this LTOP as its SCOPE argument (h5 = h1), yielding a wide scope read-

ing of the object. Since these two options are available, the example is predicted to be ambig-

uous.

A technical formulation of the Scope Principle is given in (28). Its interaction with the

HPC will guarantee that scope ambiguity may only ensue if the syntactic configuration of the

quantifiers differs from their prominence on argument structure. This can only happen if a

S(e, h

1

)|{h

1

:some(x, h

6

, h

7

),

h

6

:man(x), h

2

:every(y. h

4

, h

5

),

h

4

:book(y), h

3

:read(e, x, y)}

VP(e, h

2

)|{h

1

:some(x, h

6

, h

7

),

h

6

:man(x), h

5

:read(e, x, y)}

V(e, h

3

)|{h

3

:read(e, x, y)}

gelesen

(e, h

3

)

hat

NP(x, h

1

)|{h

1

:some(x, h

6

, h

7

),

h

6

:man(x)}

irgendein Mann

NP(y, h

2

)|{h

2

:every(y, h

4

, h

5

),

h

4

:book(y)}

jedes Buch

1

quantifier that appears lower on argument structure is syntactically realized more prominently

due to topicalization or scrambling (as e.g. in (27)).

(28) Scope Principle

The SCOPE value of a quantifier is either identical to the LTOP of its syntactic sister

or to one of the handles of its more oblique co-dependents.

2.3. Scope Options of Prepositional Complements

Coming back to prepositional complements, we have already indicated in sec. 1.3. that they

do not differ from nominal complements with respect to their scope options in German. The

pertinent examples (12) and (13) are repeated here.

(29) a. Jeder Mann freut sich auf eine Verabredung.

every man look.forward Refl on a date

‘every man looks forward to a date’ √∀∃, *∃∀

b. Auf eine Verabredung freut sich jeder Mann.

on a date look.forward Refl every man

‘every man looks forward to a date’ √∀∃, √∃∀

The analysis of (29a) proceeds along the same lines as the analysis of (26). In accordance

with the remarks made in section 1, we assume that the preposition auf in (26) is indeed gov-

erned, and semantically vacuous, and can hence be analysed as an identity function on the

meaning of its NP complement. Since the reflexive pronoun sich in (26) does not serve as an

argument, but as a marker to indicate the reflexivity of the predicate, it will not receive an

interpretation either. As was mentioned in section 2.1., we model topicalization through the

introduction of a SLASH feature. The verb in second position is also modelled through a

SLASH feature, which is called DSL (double slash), following Borsley (1989).12 So the

phrase [freut sich auf eine Verabredung] is in fact a clause that misses its subject NP, and

[sich auf eine Verabredung ø] is a VP that misses its verb. Apart from that, the analysis of

(29a) in (30) mirrors the analysis of (26) in all necessary aspects. In particular, the topicalized

subject is more prominent than the prepositional complement in terms of configuration and

argument structure, and hence, the example is predicted to be unambiguous.

(30) Jeder Mann freut sich auf eine Verabredung.

The analysis of (29b) mirrors the analysis of (27). In this case, the object has been topicalized,

and hence the subject may either take the handle of the verb or the handle of the object as its

scopal argument, resulting in two different scope options.

The analysis in (31) depicts the wide scope interpretation of the prepositional object. It

can take scope over the subject, because it is configurationally more prominent than the sub-

ject. The subject could also take wide scope over the object, since the subject is more promi-

                                                                                                               12 In sum, the present analysis of verb second is almost identical to a CP analysis with phrasal movement from inside S and

head movement from V° to C°, as in Chomsky (1986).

S(e, h

1

)|{h

1

:every(x, h

6

, h

2

),

h

6

:man(x), h

2

:some(y, h

4

, h

5

),

h

4

:date(y), h

5

:look.forward(e, x, y)}

VP/NP(e, h

2

)|{h

2

:some(y, h

4

, h

5

),

h

4

:date(y), h

5

:look.forward(e, x, y)}

VP//V/NP(e, h

2

)|{h

2

:some(y, h

4

, h

5

),

h

4

:date(y), h

5

:look.forward(e, x, y)}

VP//V(e, h

2

)|{h

2

:some(y, h

4

, h

5

),

h

4

:date(y), h

5

:look.forward(e, x, y)}

V//V(e, h

3

)|{h

3

:look.forward(e, x, y)}

DSL:V

e, h

3

PP(y, h

2

)|{h

2

:some(y, h

4

,h

5

),

h

4

:date(y)}

auf eine Verabredung

NP

sich

V

freut

NP(x, h

1

)|{h

1

:every(x, h

6

, h

7

),

h

6

:man(x)}

jeder Mann

1

nent on ARG-ST (the narrow scope reading of the topicalized object basically follows the

conditions already illustrated in (27) and has thus been omitted).

(31) Auf eine Verabredung freut sich jeder Mann.

In the following section, we will justify the claim that the same mechanism applies to adver-

bial complements and their scope options.

3. The Grammar of Adverbial Complements

As adverbial complements are indeed complements, we assume that they should be analysed

in terms of the lexical entries of the predicates of which they are complements. The analysis

must make explicit, in which respects adverbial complements function as complements, but at

the same time, the modifying nature of these complements must be accounted for. HPSG

seems to be particularly suited to carry out such an analysis, since it provides a rich lexical

structure with interwoven syntax and semantics.

S(e, h

2

)|{h

2

:some(x, h

4

, h

1

),

h

4

:date(y), h

1

:every(x, h

6

, h

3

),

h

6

:man(x), h

3

:look.forward(e, x, y)}

VP/NP(e, h

1

)|{h

1

:every(y, h

6

, h

3

),

h

6

:man(x), h

3

:look.forward(e, x, y)}

VP//V/NP(e, h

1

)|{h

1

:every(y, h

6

, h

3

),

h

6

:man(x), h

3

:look.forward(e, x, y)}

VP//V/NP(e, h

1

)|{h

1

:every(y, h

6

, h

3

),

h

6

:man(x), h

3

:look.forward(e, x, y)}

V//V/NP(e, h

3

)|{h

3

:look.forward(e, x, y)}

DSL:V

e, h

3

NP(x, h

1

)|{h

1

:every(x, h

6

, h

7

),

h

6

:man(x)}

jeder Mann

NP

sich

V

freut

PP(y, h

2

)|{h

2

:some(y, h

4

,h

5

),

h

4

:date(y)}

auf eine Verabredung

1

Adverbial complements co-occur with certain (verbal) predicates only, and are treated

as true syntactic arguments of these predicates. Consequently, an analysis of adverbial com-

plements will have to start with a presentation of the governing predicates in light of the prop-

erties mentioned in the previous sections. From the perspective of the adverbial complements

itself, we will assume that they differ in no respect from ordinary adverbial PPs, which act as

intersective modifiers. The event variable of the PP thus has to be identified with the event

variable of the modified phrase, and the PP will introduce an appropriate prepositional rela-

tion in its semantics.

As complements, adverbial complements form part of the ARG-ST of the predicates. In

addition, we must account for the fact that adverbial complements are required to show spatial

semantics. Depending on the governing predicate, further specifications are necessary. The

verb ziehen ‘to tow/to pull’ in (3) requires the adverbial PP to denote a path, while the verb

liegen ‘to lie’ in (1) and stellen ‘to put’ in (2) allow various stative local senses. The govern-

ing predicate must thus be able to constrain the semantics of its adverbial complement. In the

present analysis, this is achieved by the requirement that the governing predicate selects the

KEY of its adverbial complement and thus imposes semantic constraints on the adverbial

complement.

Still, adverbial complements provide the semantics of intersective modifiers. When

combined with their predicates, they are thus not semantically subordinated to the predicates –

they do not fill an argument slot in the elementary predication of the predication, but instead

the semantics of the governing predicates indicates the intersection. Hence, semantically ad-

verbial complements are on a par with their governors.

We will illustrate the analysis with the verbs liegen and ziehen, to provide instances of

superficially transitive and ditranstive predicates. The analysis, however, will treat them as

intranstives and transitives respectively. Formally, the analysis applying to ziehen also applies

to stellen, the only difference being the specific semantic constraints imposed on the comple-

ment. Preliminary feature representations for liegen ‘to lie’ and ziehen ‘to tow/to pull’ are

provided in (32) and (33). They still do not take the conspicuous scope variation of these

predicates into account.

(32) liegen (preliminary)

In (32), the adverbial complement PP is part of the argument structure (ARG-ST) of the gov-

erning predicate. The governor further constrains the semantics of the PP by imposing a con-

dition on the KEY of the PP, which could be interpreted alongside the traditional selection

restrictions. For the verb liegen, the condition on the KEY of the PP is that it has to be a sta-

tive locative relation (stative_loc_rel), whose LABEL and STATE variables are identical,

respectively, with the LABEL and STATE variables of the governing predicate.

The KEY of the PP is not only selected, it is also taken to form part of the core seman-

tics of the predicate insofar as the interpretation of the relation is distributed between the EP

of the verb and the EP of the P(P). That these two intersect becomes apparent through the

identification of the LABEL and the STATE variable. The intersection, however, is not ap-

2

666666666666666666666666666666664

ARG-ST

*...PP

2

6666666664

CONT

2

6666666664

HOOK 1

KEY 2

2

6666664

stative loc relLABEL h0

STATE sARG1 xARG2 y

3

7777775

3

7777777775

3

7777777775

...

+

CONT

2

6666666666664

HOOK 1

"LTOP h0

INDEX s

#

KEY 3

RELS

8>>>><

>>>>:

3

2

6664

reside relLABEL h0

STATE sARG1 x

3

7775, 2

9>>>>=

>>>>;

3

7777777777775

3

777777777777777777777777777777775

1

plied in syntax but in the lexicon.13 While adverbial complements provide spatial relations

between its internal and external arguments, the predicate itself only relates the remaining

arguments. We thus assume that liegen (‘to lie’) is syntactically transitive, but semantically a

one-place relation.

(33) ziehen (preliminary)

Similar considerations apply to the analysis of ziehen ‘to pull’ in (33). Note in particular that

ziehen will syntactically be analysed as a ditransitive verb, but semantically introduces a bina-

ry relation. The subject of liegen is thus specified only to reside statically, the subject and

object of ziehen are related such that the object is manipulated by the subject, but in both cas-

es, the spatial grounding (of subject or object) is provided by the PP. It is this tension between

the syntactic argument structure of the predicate and its content that allow adverbial comple-

ments to fulfil a double role: syntactically, they are complements, but semantically, they are

modifiers. As complements, they appear on the ARG-ST of their governors, and their gover-

                                                                                                               13 In this respect, the present analysis bears a strong resemblance to Dowty (2003).

2

66666666666666666666666666666666664

ARG-ST

*...PP

2

6666666664

CONT

2

6666666664

HOOK 1

KEY 2

2

6666664

path relLABEL h0

EVENT eARG1 zARG2 y

3

7777775

3

7777777775

3

7777777775

...

+

CONT

2

6666666666666664

HOOK 1

"LTOP h0

INDEX e

#

KEY 3

RELS

8>>>>>><

>>>>>>:

3

2

6666664

pull relLABEL h0

EVENT eARG1 xARG2 z

3

7777775, 2

9>>>>>>=

>>>>>>;

3

7777777777777775

3

77777777777777777777777777777777775

1

nors select their semantic properties. As modifiers, they provide a full-fledged intersective

semantics. The semantic contribution of their internal arguments is not directly related to their

governors. In this respect, adverbial complements differ sharply from prepositional comple-

ments, as they are usually understood.

Let us now turn to the question of scope variation. The representations in (32) and (33)

have left open the actual position of the adverbial complements on the ARG-ST of their gov-

ernors. We will assume that the basic configuration predicted by the interaction of the Seman-

tics Principle, the Handle Projection Constraint, and the Scope Principle will apply to adver-

bial complements as well, because these complements always occupy penultimate positions

on the ARG-ST of their governors. For a ditransitive verb like ziehen, this means that the PP

is found on ARG-ST before the NP, as represented in (34).

(34) ziehen

Since the PP occupies a more prominent position on ARG-ST than the object NP in (34), the

ordering of the arguments in (3) actually constitutes a case of scrambling. According to the

2

66666666666666666666666666666666664

ARG-ST

*NPx, PP

2

6666666664

CONT

2

6666666664

HOOK 1

KEY 2

2

6666664

path relLABEL h0

EVENT eARG1 zARG2 y

3

7777775

3

7777777775

3

7777777775

, NPz

+

CONT

2

6666666666666664

HOOK 1

"LTOP h0

INDEX e

#

KEY 3

RELS

8>>>>>><

>>>>>>:

3

2

6666664

pull relLABEL h0

EVENT eARG1 xARG2 y

3

7777775, 2

9>>>>>>=

>>>>>>;

3

7777777777777775

3

77777777777777777777777777777777775

1

lexical specification in (34), the PP is more prominent than the NP on ARG-ST, according to

the syntactic configuration, the NP is more prominent than the PP. The ensuing scope ambi-

guity is thus predicted. While this might be uncontroversial for ditransitives like ziehen, it is

much more controversial for transitives like liegen. According to the present analysis, the PP

is syntactically more prominent than the subject in (35).

(35) liegen

The lexical specification of liegen (‘to reside’) provides the final clue for the analysis of (1).

Since the analysis pertains to cases of topicalization and scrambling, and scrambling can be

presented more perspicuously than topicalization, which always requires the modelling of

dislocation in the syntactic structure, we will illustrate the analysis with the scrambled variety

of (1) in (36) for the wide scope reading, and in (37) for the narrow scope reading of the sub-

ject.

2

666666666666666666666666666666664

ARG-ST

*PP

2

6666666664

CONT

2

6666666664

HOOK 1

KEY 2

2

6666664

stative loc relLABEL h0

STATE sARG1 xARG2 y

3

7777775

3

7777777775

3

7777777775

NPx

+

CONT

2

6666666666664

HOOK 1

"LTOP h0

INDEX s

#

KEY 3

RELS

8>>>><

>>>>:

3

2

6664

reside relLABEL h0

STATE sARG1 x

3

7775, 2

9>>>>=

>>>>;

3

7777777777775

3

777777777777777777777777777777775

1

(36) Ich glaube, dass [ein Schimmer über jedem Bild lag]. ∃∀

The analysis in (37) differs from the in (36) only in the handles selected by the quantifiers.

The adverbial complement modifier selects the handle of the verb in (36), but the handle of

the subject in (37). The latter choice only becomes possible because we assume that the sub-

ject is less prominent than the adverbial complement on the ARG-ST of the verb.14

                                                                                                               14 Clearly, the analysis must make certain that a quantifier from a higher clause does not select the narrow scope subject

quantifier. In Kiss (2001) this is achieved through a condition that prohibits the selection of handles that have already been selected.

S(s, h

1

)|{h

1

:some(x, h

4

, h

2

),

h

4

:gleam(x), h

2

:every(y, h

3

, h

0

),

h

3

:picture(y), h

0

:above(s, x, y), h

0

:reside(s, x)}

VP(s, h

2

)|{h

2

:every(y, h

3

, h

0

),

h

0

:above(s, x, y), h

3

:picture(y),

h

0

:reside(s, x)}

V(s, h

0

)|{h

0

:reside(s, x)}

lag

PP(s, h

2

)|{h

2

:every(y, h

3

, h

0

),

h

0

:above(s, x, y), h

3

:picture(y)}

¨uber jedem Bild

NP(x, h

1

)|{h

1

:some(x, h

4

, h

2

),

h

4

:gleam(x)}

ein Schimmer

1

(37) Ich glaube, dass [ein Schimmer über jedem Bild lag]. ∀∃

Can this analysis be substantiated? The following observations suggest that the analysis is not

as controversial as it initially seemed.

The first argument comes from word order. In German, a PP preceding an NP in the

Mittelfeld is often classified as marked; this holds for PP modifiers as well as for PP comple-

ments, and might be accounted for by a general condition that prefers NPs to the left of PPs.

This condition is built on a hierarchy of categories, but other hierarchies are established as

well. One of these hierarchies determines word order on the basis of the argument’s position

on the argument structure.

With regard to transitive predicates with adverbial complements, we observe that the

order PP < NP is judged as equally unmarked as the order NP < PP, as is illustrated in (38).

(38) a. Ich glaube, dass über diesem Bild ein Schimmer lag.

I believe that above this picture a gleam lay

S(s, h

1

)|{h

2

:every(y, h

3

, h

1

), h

1

:some(x, h

4

, h

0

)

h

4

:gleam(x),

h

3

:picture(y), h

0

:above(s, x, y), h

0

:reside(s, x)}

VP(s, h

0

)|{h

2

:every(y, h

3

, h

1

),

h

0

:above(s, x, y), h

3

:picture(y),

h

0

:reside(s, x)}

V(s, h

0

)|{h

0

:reside(s, x)}

lag

PP(s, h

2

)|{h

2

:every(y, h

3

, h

1

),

h

0

:above(s, x, y), h

3

:picture(y)}

¨uber jedem Bild

NP(x, h

1

)|{h

1

:some(x, h

4

, h

0

),

h

4

:gleam(x)}

ein Schimmer

1

b. Ich glaube, dass ein Schimmer über diesem Bild lag.

I believe that a gleam above this picture lay

‘I believe that this picture was gleaming’

That word orders are unmarked might be explained by reference to the two aforementioned

hierarchies: the order PP < NP follows from the position of the arguments on the ARG-ST of

the predicate. The other order, NP < PP, follows from the general tendency to linearize PPs

after NPs. It is well known that different hierarchies may outbalance each other if word order

in the Mittelfeld is to be determined (cf. Uszkoreit 1986). This assumption mirrors the situa-

tion encountered with unaccusative predicates like unterlaufen ‘to slip in’, where the dative

argument is taken to be more prominent than the nominative argument of the predicate (cf.

Haider 2010), as can be witnessed in (39).

(39) a. Ich denke, dass dem Schüler der Fehler unterlief

I think that the.DAT pupil.DAT the.NOM mistake.NOM slipped.in.

‘I think that the pupil made the mistake’

b. Ich denke, dass der Fehler dem Schüler unterlief

I think that the.NOM mistake .NOM the.DAT pupil.DAT slipped.in

‘I think that the pupil made the mistake.’

Although the order NOM < DAT is usually the preferred one, both orders are equally possible

in (39), since NOM < DAT violates the requirement that more prominent arguments be real-

ized to the left of less prominent arguments, and DAT < NOM violates the requirement that

the nominative argument be realized in front of a dative argument.

The second argument comes again from scope considerations. The analyses presented in

(34) and (35) actually make a strong prediction concerning the scope options of examples like

(38a): if the order of arguments observed in (38a) corresponds to the order of arguments on

ARG-ST, then the HPC predicts examples of type (38a) to be unambiguous with respect to

scope. Since the configurational hierarchy corresponds to the hierarchy on the argument struc-

ture, a quantifier realized in the adverbial complement will unequivocally take wide scope of

the quantifier that is realized to the left of the subject. The same consideration applies to

scope ambiguities between the NP object and the adverbial complement in the case of ditran-

sitive ziehen ‘to tow/to pull’: there should be no scope ambiguity.

The prediction is actually borne out, as is illustrated in (40).

(40) a. Sie zogen durch jeden Abschnitt eine Schnur. √∀∃, *∃∀

They pulled through every section a cord

‘the pulled a cord through every section’

b. Es lag über jedem Bild ein Schimmer. √∀∃, *∃∀

EXPL lay over every picture a gleam

‘every picture was gleaming’

A final argument in favour of the analyses presented in (34) and (35) does not directly touch

prepositional complements, but the ordering of PP modifiers in the German clause. Frey and

Pittner (1998) argue against the assumption that basic positions can only be determined for

arguments but not for adjuncts. They propose that different modifiers may occupy basic posi-

tion as immediate sisters of the verb, and as sisters of the VP. In addition they present evi-

dence that temporal modifiers are even located in a position that is configurationally more

prominent than the position of the subject (Frey and Pittner 1998:512-516). Of particular in-

terest for the present proposal is their analysis of spatial adverbials in unaccusative predicates

(Frey and Pittner 1998:510), as is illustrated in (41) with the spatial PP in einem Park.

(41) a. Er wurde entlassen, weil ihm in einem Park wer entlaufen ist.

he was dismissed because him in a park someone went.astray is

‘he was dismissed because someone under his custody slipped away in a park’

b. *Er wurde entlassen, weil ihm wer in einem Park entlaufen ist.

he was dismissed because him someone in a park went.astray is

Frey and Pittner (1998:495) assume that weak wh-indefinites, such as wer ‘someone’, cannot

be scrambled. Hence, the wh-indefinite subject in (41a) occupies its basic position, which is

syntactically less prominent than the position of the local PP. This assumption is further cor-

roborated by the observation in (41b) that the local PP cannot be realized to the right of the

wh-indefinite. If this were possible, one could argue that the realization in (41a) is just an in-

stance of scrambling of the local PP (cf. Frey 2001:144-152).

If Frey and Pittner’s assessment is correct then (41a) provides a further instance where

PPs may occur to the left of subjects, a case that blends with the analysis of stative predicates

in (34).

4. Summary

We have presented an analysis of adverbial complements in German. Adverbial complements

are taken to be complements of certain predicates that can be characterized by a difference

between syntactic and semantic argumenthood: we have argued that these predicates select a

syntactic argument, which is not one of their semantic arguments, but instead an intersective

modifier. Hence the complements provide the full-fledged semantics of an ordinary preposi-

tional modifier, and yet are syntactically required by the predicate. In this, they differ from a

class of prepositional complements dealt with in Zifonun et al. (1997:1099-1101) as well as in

Breindl (2006): these prepositional complements are optional, and their interpretation is al-

ways laid out in the meaning of the verb.

Adverbial complements show interesting scope properties. The properties could be ana-

lysed by assuming that adverbial complements always occupy the penultimate position on the

argument structure of their predicates. For ditransitives, this implies that they take scope over

the NP object, for transitives, it implies that they are more prominent than the subject and take

scope over the subject if the subject is not scrambled across the adverbial complement.

This conclusion might be controversial at first. To substantiate the analysis, however,

we have related it to the grammar of unaccusative predicates and in addition have pointed out

that the analysis makes correct predictions for cases where no scrambling is involved.

References

ACKEMA, P. 2015. Arguments and adjuncts. Syntax – Theory and Analysis. An International

Handbook. Eds. T. Kiss & A. Alexiadou. Berlin/New York: Walter de Gruyter, 247–275.

BIERWISCH, M. 1988. On the grammar of local prepositions. Syntax, Semantik und Lexikon.

Eds. M. Bierwisch, W. Motsch & I. Zimmermann. Berlin: Akademie Verlag, 1–65.

BORSLEY, R. 1989. Phrase-structure grammar and the Barriers conception of clause struc-

ture. Linguistics 27, 84–864.

BREINDL, E. 2006. Präpositionalphrasen. Dependenz und Valenz. Ein internationales Hand-

buch der zeitgenössischen Forschung. 2. Halbband. Eds. V. Ágel, L.M. Eichinger, H.-W.

Eroms, H.J. Heringer, H. Lobin & G. Rau. Berlin/New York: Walter de Gruyter, 936–

951.

COPESTAKE, A., FLICKINGER, D.P., POLLARD, C.J. & SAG, I.A. 2005. Minimal Recur-

sion Semantics: An Introduction. Research on Language and Computation 4 (3), 281–

332.

CHOMSKY, N. 1986. Barriers. Cambridge, MA: The MIT Press.

CHRISTIE, E. 2013. Result XPs and the argument-adjunct distinction. Proceedings of LFG

2013. Eds. M. Butt & T. Holloway-King. Stanford: CSLI Publications, 212–231.

DOWTY, D. 2003. The dual analysis of adjuncts and complements in Categorial Grammar.

Modifying Adjuncts. Eds. E. Lang, C. Maienborn & C. Fabricius-Hansen. Berlin/New

York: Mouton de Gruyter, 33–66.

FREY, W. 1993. Syntaktische Bedingungen für die semantische Interpretation. Berlin: Aka-

demie Verlag.

FREY, W. 2001. About the whereabouts of indefinites. Theoretical Linguistics 27, 137–161.

FREY, W. & PITTNER, K. 1998. Zur Positionierung der Adverbiale im deutschen Mittelfeld.

Linguistische Berichte 176, 489–534.

GAZDAR, G., KLEIN, E., PULLUM, G. & SAG, I.A. 1985. Generalized Phrase Structure

Grammar. Cambridge: Blackwell Publishers.

HAIDER, H. 2010. The Syntax of German. Cambridge: Cambridge University Press.

HOJI, H. 1986. Scope interpretation and its theoretical implications. Proceedings of the Fifth

West Coast Conference on Formal Linguistics, 87–191. Stanford: CSLI Publications.

JACKENDOFF, R. 1973. The base rules for prepositional phrases. A Festschrift for Morris

Halle. Eds. S.R. Anderson & P. Kiparsky, 345–356. New York: Holt, Rinehart, and Win-

ston.

JOO, Y. 1989. A cross-linguistic approach to quantification in syntax. Madison: University of

Wisconsin. Ph.D. Diss.

KRIFKA, M. 1998. Scope inversion under the rise-fall contour in German. Linguistic Inquiry

29.1, 75–112.

MAIENBORN, C. 1990. Position und Bewegung. Zur Semantik lokaler Verben. Stuttgart:

IBM, IWBS Report 138.

MAY, R. 1985. Logical Form. Cambridge, MA: The MIT Press.

MÜLLER, S. 2002. Complex Predicates. Stanford: CSLI Publications.

MÜLLER, S. 2015. HPSG – A Synopsis. Syntax – Theory and Analysis. An International

Handbook. Eds. T. Kiss & A. Alexiadou. Berlin/New York: Walter de Gruyter, Vol. II,

939–976.

POLLARD, C. & SAG, I.A. 1994. Head-driven Phrase Structure Grammar. Chicago: The

University of Chicago Press.

SAUERLAND, U. & ELBOURNE, P. 2002. Total reconstruction, PF movement, and deriva-

tional order. Linguistic Inquiry 33 (2), 283–319.

USZKOREIT, H. 1986. Constraints on order. Linguistics 24, 883–906.

ZIFONUN, G., HOFFMANN, L. & STRECKER, B. 1997. Grammatik der deutschen Spra-

che. Berlin/New York: Walter de Gruyter.

Tibor Kiss Sprachwissenschaftliches Institut Ruhr-Universität Bochum Universitätsstraße 150 44801 Bochum [email protected]


Recommended