+ All Categories
Home > Documents > An Outline of Genericity

An Outline of Genericity

Date post: 23-Jan-2022
Category:
Upload: others
View: 3 times
Download: 0 times
Share this document with a friend
35
An Outline of Genericity Manfred Krifka partly in collaboration with Claudia Gerstner October 1987 I Universitat Tubingen, Seminar fur naturlich-sprachliche Systeme I Biesingerstr. 10, D-7400 Tubingen
Transcript
Page 1: An Outline of Genericity

An Outline of Genericity

Manfred Krifka

partly in collaboration with Claudia Gerstner

October 1987

I Universitat Tubingen, Seminar fur naturlich-sprachliche Systeme

I Biesingerstr. 10, D-7400 Tubingen

Page 2: An Outline of Genericity

In t h i s paper, I t r y t o clarify t h e notion of genericity in na tu ra l language. I show t h a t t h e phenomena considered a s generic fal l in to two qui te dis t inct c lasses , namely e i ther reference to kinds or default quantification. Default quantification, which includes habi tuals , is t rea ted a s adverbial quantification and formalized in t h e framework of discourse representat ion theory. Reference t o kinds, which includes taxonomic NPs, i s t r ea t ed by assuming kind individuals which have a n ontological s t ruc ture relat ing them t o cer tain objects, t h e specimens of t h e kind. I inves t iga te some related phenomena, a s e.g. t h e distribution of any, the accentual s t ruc tu re of generic sentences, generic anaphora and t h e interpretat ion of bare NPs. The aim of t h i s paper i s to sketch a rough map where a l l t h e phenomena considered a s generic find the i r proper place. Therefore, t he range of subjects which a re t rea ted i s qu i te broad, bu t inevi tably there a r e many a reas which a r e only touched upon and will need much fur ther research.

I profited a lo t from tile highly stimulating discussion in t h e workshop and a t the symposion on Generics, Habituals and Mass Terms, organized by Jeff Pelletier a t t he 1987 LSA Linguistic Ins t i tu te in Stanford. In par t icular , I want to thank Gennaro Chierchia, Greg Carlson, Gerhard Heyer, Claudia Gerstner, Godehard Link, Uwe Monnich, Barbara. Part.ee, Jeff Pelletier, Alice 1,er Meulen and Frank Veltman for valuable h in ts , comments and criticisms. Par t of t he ideas in t h i s paper were developed in colloboration with Claudia Gerstner for a n ar t ic le t o appear i n Handbuch der Syntax, ed. Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann.

T a b l e of C o n t e n t s

1. Types ot Genericity 1.1 A Reference Classification of Genericity 1.2 The Two Underlying Types of Genericity

2. I-Genericity, or Default Q~antificdtion. 2.1 I-Generic Sentences 2.2 Habituals

3. P-Grinerlcity as Reference to Kinds 3.1 Non-taxonomic D-generic NPs 3.2 Taxonomic NPs 3.3 The Meaning of D-Generic Sentences

4. Genericity Recl.assified

5 . Some Related Phenomena 1 I-genericity and Downward Entailment: The Case of "Any" 5 . 2 Set P i edication 5.3 Do We Need Complex Model Structures? 5.4 Accentuation, The~neIRheme-Structuring, and Quantifiers 5 . 5 Generic Anaphor a 5.6 Generic and Episodic Interpretation of Bare NPs

References

Page 3: An Outline of Genericity

1. TYPES O F GENERICITY p-

1.1 A Reference Classification of Genericity

I will s t a r t by giving a classification of generic NPs. There i s no theoret ical claim in th i s classification; I will even show t h a t a uniform notion of a "generic NP" i s not a well-founded one. The sole aim of t h e classification i s t o serve a s a reference system for l a t e r discussion.

1.a) The lion is a ferocious beast. (singular definite generic NP) b) The lions are ferocious beasts. (plural definite generic NP) c) A lion is a ferocious beast. (singular indefinite generic NP) d ) Lions are ferocious beasts. (generic bare plurals) e ) Gold is a precious metal. (generic bare singular) f) One cat, namely the lion, is a ferocious beast.

Some cats, namely the lion and the tiger, are ferocious beasts. ( taxonomic NPs )

[g) John smokes. (habitual sentence) 1

Habitual sentences have been t reated a s generic a t l ea s t s ince Chafe (1970) and Lawler (1973), although there i s nothing l ike a "generic NP" in them. B u t t he re i s actual ly an import,ant theoretical insight in grouping habi tuals together with a t least one so r t of generics. I will come back t o t h i s issue la te r .

If we look a t t h e examples given above, i t becomes immediately clear t h a t genericity cannot depend on t h e syntact ical or morphological form of the NP. Every "generic" NP tiled so far could occur a s a non-generic NP in o ther contexts . And a s f a r as I can see, no language marks exactly those NPs which a r e t o be interpreted as generic.

B U L there a r e NPs which can only be interpreted a s generic. For example, t h e English IMP (not, t he common noun) Inan can only be interpreted a s generic, and the re are some markers for taxonomic NPs, like kind, or German Art , Sorte.

! iu Man h a s lived in Africa fo r more than 2 million years. b ) Some kinds of cats live in t he bush. r ) Kiriifie KatZerkarLen leben in der Steppe.

1 & H I come back l.o those NPs in due course.

1 h u reference classif~c 'a t ion 01 generic NPs presented in (1 ) could indicate t h a t t i n - r u ib a ht-iwilck'ririg cornple~i ty with generic NPs. I will show t h a t , on t h e ' o n t r a r y , b e have to distinguish only two kinds of genericity, To show th is , I will i i o k a t tlie distribution of t he types of generic NPs with respect to different c lasses of predicates. 1 will look a t four diagnostic contexts.

W t ? ill start, with considering kind predicates l ike be extinct, which can be analyzed a s having the selection restriction that, t h e subject must be a kind. We observe t h a t s(:tilt;nres with indefinite singular NPs do not pass t h i s test,.

Diagnostic context ( i ) : Kind predicates 3.a) The lion i s ext inct .

b ) The lions a r e ektirlft . c ) * A lion is extinct,. d ) Lions a r e ext inct . e ) Bronze was invented before 2000 l3.I" f ) A ca t , namely the lion, is extinct..

Page 4: An Outline of Genericity

Secondly, we will consider dynamic predicates (i.e. non-stat ive predicates). Again we note t h a t sen tences with indefinite s ingular NPs do not pass t h i s t e s t . Their verbal predicate must be s t a t i ve , whereas i n o ther cases , t h e verbal predicate may be dynamic ( a s observed for definite generic NPs by Heyer 1985).

Diagnostic context (ii): Dynamic predicates 4.a) The r a t reached Australia in 1770.

b) The rats reached Australia in 1770. c) *A r a t reached Australia i n 1770, (generic reading) d) Rats reached Australia in 1770. e ) Rice was introduced in East Africa some centuries ago f ) A rodent (namely t h e rat) reached Australia i n 1770.

Thirdly, a s noted by Lawler (1973), some generic NPs cannot combine with predicates expressing accidental properties, but only with predicates expressing properties which a r e in some way essential for Ihe referent of t h e generic NP. Since dynamic properties a r e always accidental, t h i s explaines why indefinite generic N P s cannot combine with dynamic verbs. A s t he a t t en t ive reader will a l ready expect , singular indefinite NPs do not pass t h i s t,est.

Diagnost ic contex t ( iii): Accidental p r o p e r t i e s (Lawler) 5.a) The madrigal is popular. a ' ) The madrigal is polyphonic.

b) The madrigals a r e popular. b') The madrigals a r e polyphonic. c ) *A madrigal is popular. c ' ) A madrigal is polyphonic. d ) Madrigals a r e popular. d ' ) Madrigals a r e polyphonic. e ) Music is popular. e ' ) Music is polyphonic. f ) A type of music (namely, t h e i') A type of music (namely, t h e

madrigal) i s popular. madrigal) i s polyphonic.

Final ly , w e will look a t the generic N t J i t self . Vendler (1967), Nunberg & Pan (1975), ~ a r l s o n (1982) and Uahl (1985) pointed out t h a t no t every noun allows for a definite singular generic NP. Basically, the noun must be semantically related to a well- established kind. This Lest blocks sentences with definite singular and plural NPs and sentences with taxoriomic W s .

1)ia~:iiostic context, ( i v j : Well--est,ablished k inds 6 . a ) The lion is frirodous

1 ) ) The lions a r e ferocious. f ) A lion is ferocious. d l Lions a r e feroeiuiis. e ) Gold is precious. t ' i A rat ( n a m e l y t h e lion) is ferorious.

'1 a ) *The lion wi th th ree legs i s ferocious. I)) '?TIie liuns with 1-hree legs a r e Ferocious. e ) A lion with three legs is ferocious. d ) 1,ions with threo legs a r e ferocious. e ) Gold which i s hanimcred f l a t is precious. f ) ' A kind of lion (naniely t h e lion with three legs) is ferocious.

T t i ~ reason why (7 .b) is not completely bad i s not relevant for t h e quest ions we a r e ( - ' i ~ i i ( ' t ? f ~ ~ f ? d with here: the lions with three legs can be analyzed a s referring t o t h e .sum individual of all t h e lions with three legs, bu t i s then not a proper generic NP. The difference can he easily seen if a kind predicate is involved, which cannot be applied to individual sums.

8.a) The lions a r e extinct b ) *The lions witli three li-'gs arc nxtincl

N o w let us eva lua te Hie four tes l s . Tests ( 0 - ( i i i ) f i l ter out singular indefini te generic NPs, and t e s t ( i v ) f i l lers out definite generic and taxonomic NPs. I will

Page 5: An Outline of Genericity

call NPs which pass t e s t (i)-(iii) D-generic because t h e most typical representa t ives of them seem t o be defini te generic NPs. NPs which pass t e s t ( iv ) I will call I- generic, alluding t o indefinite generic NPs.

gener ic NP D-Generics I-Generics

s ingu la r d e f i n i t e X p l u r a l d e f i n i t e X s ingu la r i n d e f i n i t e - bare p l u r a l X bare s ingu la r X taxonomic X

We see t h a t two types of generic NPs, bare plurals and bare singulars, occur in both classes. One of them, bare plurals, ha s been t h e main focus in t h e semantic analysis of genericity, in t h e t radi t ion of t h e classical analysis of Carlson (1977). I think t h a t , in hindsight,, t h i s was not t he most felicitous decision. If we want t o develop a semantics for genericity, we should concentrate a t f i r s t on those NPs which occur only in one class , tor example singular def ini te generic NPs and singular indefinite generic NPs, and then t ry t o extend the analysis t o cover NPs occuring in both classes.

I will argue t h a t D-Generics and I-Generics a r e in fac t qui te different syntact ical ly a s well a s semantically. In shor t , I will analyze D-Genericity a s reference t o a specific type ol' individuals, namely reference to kinds, which basically is NP- oriented. I-genericity, on t h e other hand, will be t r ea t ed as defaul t quantification. whii'h lias scope over t he verbal predicate a s well. Let u s s t a r t by looking a t I- Genericity.

'. 1 1-generic sentences

. t ; i u s first iouk at lhib type ol' NPs which occur a s 1-generic NFs. The basic rule here is thai every indefinite NP can occur a s a n I-generic NP. By "every indefinite N C " 1 h; ive in mind every syntact ic type of indefinite NP (count noun with ar t ic le , bare plural, bare mass term), a s well a s NPs which a r e deliberately complex, like l i i - ~ r t s with three &s This is shown by t h e following example.

9 . c There i s a lion (with three legs) in the cage. b) Th~i - t - t a r e hens (with three legs) in t,he cage. c ) Tiirire is gold (which is hammered f l a t ) in t he safe .

Secondly, we found ~ h a l t he verbal predicate in 1-generic sen tences expresses a n essent ial property. A s a verbal predicate l i k e is ferocious does not necessarily : ' <p res s an essent ial property by i tself , one h a s to assume a n operator with scope over t,he verbal predicate which expresses t h i s essent ial i ty .

' h e r e is m o t her reason why we should assume t h a t t h i s operator h a s scope over t h e verbal predicate. In m a n y languages, I--generics can be formally marked by some device (which is used to mark tiabiluals a s w e l l ) . Now, t h i s marker i s e i ther some auxi l iary (as in t h e English u s e d to-construction or t he corresponding German pf leg t zu- construct ion) , some adverb (like Knglish usually), or an affix of t h e main verb, a s i n Swahili, Japanese, Turkish and many other languages. I give two examples, one from German and one f rom Swahili.

Page 6: An Outline of Genericity

Ein Loewe pflegt gefaehrlich zu se in ' a lion uses dangerous t o b e A lion usual ly i s dangerous'

Mtoto hu-penda micheza. child HAB-love plays 'A child loves games'

But we never find a case where t h e marker of I-genericity i s pa r t of t h e NP. This holds a t l ea s t if we exclude adjectives l ike typical in a typical lion, which do not s t r ic t ly mark I-generic NPs, because they occur also with non-generic NPs, e.g. in predicative NPs a s in Simba is a typical lion. The syntac t ic position of t h e markers of I-genericity, then , shows qui te clearly t h a t I-genericity involves t h e verbal predicate.

Our basic findings concerning I-genericity can be summarized a s follows:

i ) Every NP which can serve a s a n indefinite NP can also serve a s a n I-generic NP, in a n appropriate context .

i i ) I-generic sentences have a n operator with scope over t he verbal predicate, expressing t h e notion t h a t t h e property of the verbal predicate i s a n essent ia l one for t he subject referent ,

The analysis of I-genericity now seems t o be r a the r obvious, given t h e theories of indefinite NPs of Kamp (1981) and Heim (1982). As i s well known, Kamp and Heim proposed theories in which indefinite NPs a r e not quant if iers , bu t predicates which get the i r quant if icat ional force by other quantificational elements. Consider t h e following example in Heim's analysis:

11) Often, if Pedro has a donkey, he bea ts i t . If Pedro has a donkey, he often bea ts i t . MANY ([have ' ( x,y) , donkeyl(y), x=Pedro'l,!beat'(x,y)l).

l i ew, M A N Y is a universal quant if ier which has two arguments, t h e antecedent and t h e consequent It can be interpreted as : Many cases which have x ,y meeting t h e sti~lecedeitl are such t h a t x,y meet the consequent a s well. The quant if ier can be specified by adverbs l i k e always, often, rarely, never, or i t can be le f t unspecified a.ix.1 is in t h i s case interpreted a s some so r t of universal quantification.

\2) I f Pedro has a donkey, he always/often/rarely/never/O beats i t .

i w ~ o r e w e analyze 1-generic sentences in more detai l , l e t u s look a t some general principles of quant if ier interpretat ion in t he s ty le of Kamp and Heim, Quantifiers h a w the followiiig f'orni in t h i s theory:

Here, xt . .x j a re t he variables of t h e an tecedent par t , and xn..xm a re t h e variables of t h e consequent p;irl which a r e bound t,o t,he quant if ier QUANT in Heim's theory. In Kamp's theory, xi..x,j a r e t he discourse referents introduced in t h e an tecedent par t , and X n . . X m a r e t he discourse referents introduced in t h e consequent pa r t , respective- ly. Essentially, X I . . X ~ and xn..xm a r e free in t h e antecedent and consequent par t . But antecedent and vor~sequent, inay contain some more free variables, which a r e bound by operators having wider scope, and th i s makes i t necessary for t he variables which a r e actual ly bound t o Ilie quant if ier t,o be coiniiexed on them (here, I l i s t t h e antecedent var iables and the consequent var iables in square brackets following t h e quant if ier) . X and Y is a convenient shorthand notation for t h e s e t È variables xi . .x j and XU..XB, respectively Normally, there is some overlap in var iables between antecedent and consequent, t h a t is, XnY#?b. QUANT i s associated with a number or a cer tain proportion, e .g . most is associated with "more than half". The formula (13.a)

Page 7: An Outline of Genericity

t hen can be interpreted approximately as:

13.b) In QUANT variable assignments a , for which ANTECEDENTW] is t rue , CONSEQUENTIY 1 i s true.

But t h i s only holds if Y i s a subse t of X. If t he re a r e var iables i n Y which a r e not in X, t hey must be exis tent ial ly quantified. This amounts t o t h e following:

13.c) QUANT[X;Y~(ANTECEDENT(X],CONSEQUENT~Y 1) <=> For QUANT variable assignments a , for which ANTECEDENTIX] is t rue , t he re a r e var iable assignments a' which a r e ident ical t o a with t h e possible exception t h a t they differ in t h e assignment for var iables in Y\X, for which CONSEQUENrI'lYl i s t rue .

For example:

14) Every lion h a s a matie. EVERY 1x;x.y j((lion'(x)J, Imanel(y) & have'(x,y)]) = Wxlliont(x) -> 3y[mane1(y) & havel(x,y)]]

This can be read as: For every assignment, a , for which lion'(x) is t rue , t he re i s a n assignment a' identical t o a with t h e possible exception t h a t a ' (y) F a(y) , for which manel(y) & havel(x,y) i s t rue . This amounts t o t h e first-order formula given above.

N o w , let us re turn to I-generic sentences. They can be analyzed in a similar way to quantified sentences: The so-called "generic" NPs a r e just indefinite NPs without qi-iantificational force of the i r own, and t h e quant if ier can be associated with t h e verbal operator we noticed with I-generic sentences. This becomes even more obvious if we consider t he Fact t h a t I-generic sen tences resemble sentences with over t adverbial quant if iers ( a point already made by Biggs 1978, who t races genericity back l o ;i quarttificational adverbial which is not expressed overtly; for adverbial quan t i l~ ( - aHou in general el'. Lewis 1975;:

l 5 . a ) A l ion has a inane. b ) A l i u n always/iriostiy/often/seldoni/never has a mane.

i n Ian., exact ly t h i s theory was proposed by Heim (1982). There a r e somewhat similar ai-a(soiirits h y Parkas & Sugioka ( 1983) and Pelletier & Schubert (1987). Furthermore, t h i s theory captures t he intui t ion t h a t a n I-generic NP does not refer t o a pariicular refereni , t h a t i s , i s not bound to a n exis tent ial quant if ier (cf. Jarkarniofi 1972).

i f ttu?rc is ail over1 adverbial quant if ier , t he analysis seems t o be qui te clear. (Jonsider t he following example, which can be paraphrased as: in many cases , if x meets 1 he an tecedent , i .e. i s a lion, the11 x meets the consequent, i .e . i s ferocious.

16) A lion i s often ferocious. MANYlx;x~([lion'(x)],[ferocious'(x)~)

But which quant if ier should we propose for I-generic sen tences lacking a n over t quantifier? It cannot be the universal quant if ier , given t h e well-known fac t t h a t 'except ions" are admissible with generic sentences. There have been some proposals in recent years l,o assume instead an operator based on defaul ts , e.g. Plat teau (19801, Strigin (1985), Heyer (1985), Garlson (1986) and Pelletier & Schubert (1987), or t h e similar notion of s tereotypes, e .g . Geurts (1985). 1 think th i s i s on t h e right t rack for t he ana lys is of I-generics, but t h a t t h i s defaul t operator should be analyzed a s a quant if ier in t he :sense of Kamp and Heim. 1 will simply ra i l i t DEFAULT.

Here, I am not going t o t a lk much about t h e semantics of t h i s defaul t quant if ier . Rather informally, i l can be given a s follows:

Page 8: An Outline of Genericity

17) DEFAULT[X;Y](ANTECEDENT[X],CONSEQUENT[Y)) :<=> For a n y variable assignment a, for which ANTECEDENTtX] i s t rue , there is a var iable assignments a' which is identical t o a with t h e possible exception t h a t a and a' differ i n t h e assign- ment for var iables in Y\X, for which CONSEQUENT(Y1 i s t rue , except when CONSEQUENT[Y] ( re la t ive t o a') is not consis tent with t h e fac ts assumed so far .

The difference t o o ther quant if iers i s t o be found in t h e last clause which res t r ic t s t h e universal quantification to cases in which the consequent i s compatible with the explicit or derived information assumed so far . Of course, t h i s presupposes t h a t t h e fac ts assumed so f a r can be identified in some way. Basically, defaul t quant if i - cat ions a r e defaul t rules as described by Reiter (1980). They should be handled in some kind of non-monotonic logic, for example in d a t a semantics a s proposed by Veltman (1987) . Here, I simply want t o make clear t h a t t h e defaul t operator we need should have t h e form of a quant if ier i n t h e Kamp/Heim analysis.

Let u s consider th ree puta t ive counterexamples which show how the definition of DEFAULT i s intended to work.

18.a) A bird lays eggs. b) A t u r t l e grows very old.

One could argue t h a t only female birds lay eggs, and t h a t t u r t l e s only rarely l ive to be very old s ince nearly a l l of them a r e killed by predators when s t i l l young. But these objections can be rejected if one bears in mind t h e restr ic t ion in t h e definition of t h e defaul t quant if ier . In (18.a) , t h e inference i s blocked if among the f ac t s assumed so f a r a r e the propositions t h a t only female animals give bir th , t h a t laying eggs is a form of giving bir th , and t h a t birds a r e animals. In (18.b), t,he inference i s blocked if t h e assumed fac ts contain, among others , t h e proposition t h a t t h e life of a n animal can be shortened by various external reasons. This shows t h a t there is a lot to b e expected from pragmatics, background assumptions and background inference to a r r ive a t a correct interpretat ion of generic sentences.

Fhure is another difference between DEFAULT and other adverbial quant if iers on the u 1 1 ~ hiiiia and nominal quant if iers a l t he other . Nominal quant if iers can be i inrus~nct ,ed or restr ic ted t o some given domain (cf. 19.a,b). But t h i s i s no t t he ease wi th t he DEFAULT quant if ier (cf. Croft 1986 for t h i s observat ion) and for adverbial quantification in general: they always a re unrestr ic ted (c f . l 9 . c ,d ) .

19 .a ) ( O u t o f tho blue:) Every lion has a mane. (non-restr ic ted) b) There a r e lions arid t igers in tku: cage. Every lion has a mane.

( restr ic ted or non-restricted) C - ) 1lit+re a r e 1 1 ~ ) ~ s and tigers in t hc cage. A lion has a mane.

([ion-restricted only) d ) 1'here a r e lions and t igers in t h e cage. A lion always h a s a

inane, tnon-restricted only)

A fur ther important point i s t ha t t h e DKFAULT quant if ier can employ different "modal dimensions". (Thai i'enericity i s a modal concept has already been observed by Dahl 1975, of course) A s often with modal operators, language does not specify the modal dimension. Wit,h 1-generic sentences, i t can be linguistic necessity or analyt ic i ty a s in (20.a), n ia t t~ - '~~ ia t i ca l necessity a s in ( b ) , objective necessity a s i n (c) arid in t h e examples we considered so far , deontic necessity a s in (d), and even a so r t of ne- cessi ly which is to be relativized t o t i m e and space, a s in (el .

2O.a) A sp ins te r is unmarried. b) Two and two equals four. c) A lion has a mane. d ) A boy doesn't cry. e ) Three kiwis a r e sold for one dollar

Page 9: An Outline of Genericity

f) A unicorn h a s a horn.

The l a s t example, (f) , shows t h a t a generic sentence can be t r u e even if t he re i s no individual in t h e rea l world which sa t i s f ies t h e antecedent , a f ac t observed e.g. by Carlson (1982). The DEFAULT quant if ier simply s a y s t h a t if a n en t i t y meeting t h e antecedent ex is t s , t hen t h e consequent applies t o i t by defaul t . Therefore, defaul t quantification can be shown t o be equivalent t o quantification over possible objects. This shows once more t h a t I-genericity has t o be handled a s a modal notion. But here, 1 will not go in to t h e involved question how t o model modality, e.g. with possible worlds or some other devices.

Related t o t h e modal dimension i s t h e s t r ic tness of t he generic predication, t h a t is , whether exceptions a r e admissible or not. Linguistic necessity and mathematical necessity do not allow exceptions, whereas o ther modal dimensions should allow them in principle.

The analysis of I-genericity can be extended to other kinds of indefinite NPs a s well, e.g. t o plural NPs like three kiwis, t o bare plurals l ike lions, and t o bare mass terms like gold. provided t h a t a proper semantics of indefinite NPs of these types i s given, a s e.g. in Link (1983), Krifka (1987).

As an interest ing and well-known example, l e t u s consider sen tences l ike unicorns h a v e horns. The plural of horns was sometimes considered t o be a syntac t ica l dependent plural, defying a compositional analysis (see the discussion of unicycles have wheels in Chornsky 1975). Here, I want t o show t h a t they can be t r ea t ed cornpositionally. Suppose we can handle collective (or be t te r , cumulative) and dis t r ibut ive readings for non-generic sen tences a s in t he following examples along the lines suggested by Link (1983). To make t h e argument shor t , I simply assume a dis t r ibut ive operator on the verb t o handle t he dis t r ibut ive reading, whereas t he bare v e r b should cover t h e cumulative reading. x i s supposed t o be a sum individual roiisisLing of some contexlually sa l ien t unicorns, have-a-horni i s a predicate applying t o en t i t i es which have exactly one horn, and have-horns' i s a predicate applying t,o enl,ities having more than one horn.

2,l .a) These uniroriis have a horn x=these-unicorns' , DlSTR(have-a-horni)(x)

b ) These unicorns have horns x-these-unicorns', have-hornsf(x)

(a i says t h a t i t holds distributively for x, t h a t x have a horn (and by a proper acf ' ir~it iun of I~IS'I'R, this ineaiis t h a t every atomic par t of x has a horn) . (b) s ays t h a i i t holds cumulatively for x, that. x have horns (and t h i s would be t rue if , e.g., A c o n s i s t s of three unicorns). Given th is analysis of dis t r ibut ive and cumulative readings, i t b e e o i r ~ t ~ h tr ivial lo ;icconnt for t h e examples (22). Here, unicorn' and unicorns' a re t u be interpreted as predicates applying t o single unicorns and indiv iauals consisting of several unicorns, respectively.

22.a) A unicorn has a horn. ~~~~~~~,TIx;xl(~unicorn'(x)l,[have-a-horn~(x)l)

b ) Unicorns have horns. DEFAULT[ x;x ](I unicornsi(x) 1, lhave-hornst(x)])

c ) Unicorns have a horn. u~~~~~~~x;xl((unicorns'(x)],~UlS'~l~~have-a-horn')(x)~

We get reasonable possible interpretat ions for (22.a,b,c), a s shown by t h e formalizations given above.

The analysis developed here r a n easily be extended Lo cover atemporal when-clauses, a s discussed by Carlson (1979) and Farkas & Sugioka (1983).

Page 10: An Outline of Genericity

23) Lions a r e ferocious when they have red eyes. DEFAULT(x;x]([lions'(x), have-red-eyesf(x)],lferociousf(x)l)

Here, t h e when-clause simply adds more restr ic t ions t o t h e antecedent par t . Note t h a t a n analysis along these l ines must assume a fair ly free mapping from syntac t ic form to semantic representation. In section (6.4) of t h i s paper, I will show t h a t there a r e intonat ional cues which could be exploited t o assemble t h e pa r t s of t h e sentence going to antecedent or consequent of t h e quant if ier . But surely, more work needs to be done t o determine t h e possibilities for t h e mapping relation between syn tax and semantics.

I t has of ten been observed t h a t generic sentences and habi tual sen tences share common character is t ics (cf. e.g. Chafe 1970, Lawler 19731, and they have been t r ea t ed a s essent ial ly t h e same in formal analyses l ike the ones presented by Dahl (1975) and Carlson ( 1977 1.

There a re several reasons why habi tuals and I-generics should be t rea ted along the same line. Firs t , t he verbal predicate in 1-generic sentences and habi tua l sentences must be s t a t i ve . This explains why a progressive predicate (which must be episodic in s tandard ana lyses) cannot give rise to a habi tual interpretat ion:

24.a) John smoked. (habi tual or non-habitual) b) John was smoking. (non-habitual only)

Actually, (b) can have a generic reading a s well, a s pointed out by Chierchia (p.c.1. B u t I think th i s simply reflects the fact. t h a t many s t a t i ve predicates can occur in the progressive in certain circumstances.

Secundiy, i t is of ten t h ~ cast! t h a t explicit markings of habi tual i ty (like t h e tinglish used to-construction, t he German pflegt zu-construction, and t h e Swahili verbal marker hu-1 occur in I-generic sent,ences a s well:

:;!;.a) tiarks pt'legt Zigaretten zu rauchen. H; i i i s uses cigaret tes to smoke

1 ) ) K i n Loewe pt'legt eine Mahne zu habe r~ . a lion USPS a mane t o have

Furlhermort;, h a b i t u a l sent,ences a r c relat,ed to sentences containing quantificational adverbs, just a s sentences with adverbial quantification a re .

:& ) John a1 ways/ufi (?n/usually/rarely/never/D smokes

The~'el'o!'~, habi tual i ty should be t rea ted like I-genericity, i.e. as defaul t qu;irit,i<'i~;at,ion. Th is becomes possible i f we introduce occasions, a n idea present in ,awl(Â¥* ( 1 972, 19731, Spears (1974), Newton (1979), Kleiber (1985), Pelletier & Schubort (1987). This c o u l d be easily done hi frameworks like even t semantics or s i tuat ion semantics, which provide u s with space-time locations in t h e i n terpretal Ions of st:ritences.

I'tie semantics 01' habi tuals c an be captured most. easily when we s t a r t with a slightly more complex example.

27) John smokes a f t e r dinner. DEFAULT! \ , I ;x, t l(lx=John' , after-dinner'(t,)l,lsrnoke'(x,t)l)

This s ays that, fo r every x, x=John, iind t , a n occasion af te r dinner, x smokes a t t by

Page 11: An Outline of Genericity

default , t h a t is. if there i s no explicit reason why he does not smoke (e.g., he may have run out of cigarettes). Note t h a t the prepositional phrase se rves t o res t r ic t t he antecedent (l ike the when-clause discussed above), and secondly, t h a t the verb smoke h a s t o be relativized t o occasions.

This relat ivizat ion of dynamic verbs t o occasions seems t o be necessary in other respects as well. For example, Partee (1973) discussed negations of expressions containing dynamic verbs, which clearly refer t o a specific occasion. For example, (27') does not mean t h a t John never turned off t h e s tove, bu t he didn't t u r n i t off on a specific occasion. Therefore, t h i s sen tence has t o be formalized not by (a ) , bu t by (bl , where PRAG should single out a pragmatically sa l ien t occasion.

27') John didn't t u r n off t h e s tove. a ) -I turn-off'( John', the-stove') b) 3t[PRAG(t) & -I turn-off ( Johnl , the-stove ' , t ) J

How should we analyze simple habi tual sentences? The most na tura l t reatment is , I think, one which follows t h e pa t te rn of complex habi tual sentences. We assume t h a t with simple habi tual sentences, t he antecedent does not specify tlie occasions explicitly, but leaves th i s task to pragmatics (cf. Spears 1974, Newton 1979 and Kleiber 1985 for similar assumptions). These pragmatic restr ic t ions a re , however, notoriously difficult t o s t a t e . I think t h a t our example can best be characterized as: there a r e conditions under which John smokes by default . This yields t h e following representation:

Formally, 1 introduce a predicate variable P, for which I claim t h a t i t holds a t least, for some occasions, and which i s bound by a n ex is ten t ia l quant if ier ranging over t h e whole sentence (note t h a t t h i s exis tent ial closure i s necessary for t h e hainp! Hoiin Lrt:aUnenI anyway ).

in order to obtain ri uniform analysis , t h e variable P should be assumed in complex habiLna.1 sent.ences a s well, a s in (291, In t h e following, however. I will s t ick t o I t i(+ si i f iplu forniaii?:ition.

29) John binokes a f te r dinner. 31'1 l)I-';PAl il.TI <,L;x,t l ( [x= John' , P=after-dinner', P(t.) j , [smoke'(x,t)l)j

1 1 is i n l e r e s t i r i ~ 1.0 look a t t h e way negated habi tual sentences a r e handled in t h i s approach . I assume t h a t negation can have wide or narrow scope. Then we get t he Following analyses:

3o.a) ,lot111 doesn't smoke a f t e r dinner i ) -, L)EP'AIJLTI x,t.;x, tKlx=John', after-dinner'(Dl.lsrnoke'(x,t)!) l i ) i)EFAlil.I~~x,t,x,t~(~x=John', after-dinnerl(t,)l,[-i srnoke'(x,t)])

(30,a . l ) can be p a r ~ p h r a s e d as : There is no defaul t t h a t John smokes a f t e r dinner, and OO.a.ii) as: There i s a defaul t thal, John doesn ' t smoke a f t e r dinner. This captures t h e readings of (a) quite well. (30.b.i) can be paraphrased as: there a r e no conditions F' under which there i s a defaul t t h a t John smokes. This gives t h e se- nianticb oi negated simple habi tuals . (30.b.i i) can be paraphrased as: There a r e conditions under which there is a default t h a t John doesn't smoke. That (30.b) intui t ively lacks th i s reading can be explained by i t s very weak t r u t h conditions, which make i t useless for human eomrnunication. For example, there a r e a lot of conditions, e.g. when John i s asleep, under which John doesn't smoke by defaul t .

Page 12: An Outline of Genericity

I leave here t h e interest ing subject of negated habi tuals t o make two remarks on my approach t o habi tua ls i n general.

The f i r s t point i s t h a t a semantic character izat ion of habi tual i ty like t h e one developed here may be considered unsat isfactory, because i t relies on a lo t of pragmatic considerations. But I think t h a t defaul t quantification simply draws heavily on background knowledge. I think t h a t we cannot be more specific in semantics t han jus t t o pinpoint t he spots where pragmatics creeps in to semantics i n cases like th i s .

Secondly, note t h a t t h i s analysis can explain why habi tual predicates a r e s ta t ive . We can analyze non-stat ive, episodic predicates a s predicates which have a n occasion argument t h a t i s filled by the ex is ten t ia l quant if ier ranging over t h e t e x t ( the exis tent ial closure). They give a report on a specific event or s i tua t ion , a s i t were. Look at t h e following example:

3 1 ) (...) John smoked. (...) 3x, t ,... [ (. , .), x=Johnn, smokel(x,t) , (.. .) I

If t h e predicate lacks a n occasion argument filled by a n occasion variable which is bound by th i s exis tent ial quant if ier , i t must be considered a s s t a t i ve . There a r e two cases in point: e i ther t he verbal predicate i s inherent ly independent of occasions. a s e.g. have a mane, know French, or the occasion variable ge ts bound by some quant if ier other t han the exis tent ial quant if ier ranging over t h e whole t ex t . In botn cases , t h e predication i s independent of specific occasions. I t seems na tu ra l t o equate t h i s independence of specific occasions with s t a t i v i ty . But then , i t follows that default, quantification implies s t a t i v i ty , t h a t is , why I-generic sen tences and habi tual sentences have s t a t i v e verbal predicates (see Newton 1979 for a similar explanation why LMe predicates in those sentences a r e imperfectrive in Modern Greek). Of course, s t a t i ve verbs like knou French must be interpreted re la t ive t o a time - we iiave pi ' ist , present and fut,ure Forms -, but t h e noiion of occasion int,roduced above s h o u l d be considered a s different Ironi the notion of l ime).

:l I i\on-Laxonurni( 1)-Generic NPs

i w i i i n o w turn to 0--gerlericiLy, t he other type of genericity we have identified. The b a s i c i ~ i d i n g s about U-genericity ran be recapi tulated a s follows:

I or* predK-at es wnich a r e associnl ed with a n established kind can ynl~ii a l)-generif NP .

1 1 ) 11 is not. required t h a t t he verbal predicate i s s t a t i ve , or in 'A special mood, like i t is with 1-generic NPs.

i i i ) Tilere a r e soin? verbal predicates (kind predicates) which require a 0-generic NP in sonic argument place.

'i'r:(\,ie t':icts can be must, easily explained, of course, if we assume t h a t D-generic NPs r ~ f e r Lo a special sort of en t i t i es , namely kinds. This means t h a t D-geiiericity i s basically a n NP-orientfed phenomenon, unlike I-genericity, which we have found t o be sentence-oriented. Observation ( i i i ) , which might suggest t h a t D-genericity i s verb- oriented, can be explained 'AS a sor ta l st\ lerlional restriction of t hese predicates, similar l o t he select i ona l restriction t h a t t he object of t h e verb drink must refer to a l iquid,

1 will first examine non-t.axoriomic NPs. In Knglish, these NPs a re typically definite

Page 13: An Outline of Genericity

singular NPs like the lion, bare plurals l ike lions, bare mass terms like gold and, probably somewhat less felicitous, definite plural NPs like the lions. In t h i s section, I will t r y t o motivate t h e forms of D-generic NPs we find in English and other languages.

Kinds should be construed a s individual en t i t i es , and D-generic NPs consequently a s expressions referring t o those en t i t i es . The subclass of D-generic NPs we consider in t h i s section should furthermore be analyzed a s proper names (cf. Langford 1949, Carlson 1977, Heyer 1985). One t e s t i s t h a t they can be t h e subject of t h e so-called so-called construction, in contrast t80 ordinary definite descriptions, a f ac t Carlson (1977) observed with bare plurals.

31I.a) Big Joe i s so called because he is real ly f a t . b) *This man i s so called because he i s really f a t . c ) The liger i s so called because i t is t h e offspring of a lion and a t iger .

There a r e some lexical expressions which can be analyzed directly a s proper names of kinds, for example the English NP (not t he noun) man and scientific names of kinds, like homo sapiens.

32.a) Man lived in Australia for a t l ea s t 40000 years . b) Homo sapiens lived in Australia for a t l eas t 40000 years .

Hut. normally, nouns have a t l eas t two functions: on t h e one hand, they a r e related to a n individual en t i t y , t h e kind. and on the o ther , they a r e related t o every object which belongs to t h i s kind, t h a t i s , to every realization of t he kind. According t o the first, function, they shuuld behave like proper names, and according t o t he second one, they should behave like predicates.

of course, these two meanings a re clohely related to each other . An intui t ively appealing inoaol is t h e following: Basically, a common noun denotes a kind. Suppose we have a reialiuii R which holds between kinds and the i r realizations (cf. Carlson 9 7 7 ) Then :t mass noun like gold, which refers t o a kind Au, also has t h e s e t of I-ea1~z:itions 1 ) i this k ina , ^.xlIt(Au',x)l, a s a n extension. The same holds for bare p lu ra l n o u n s like ln'ns: they denote a kind, leo, and the s e t of real izat ions of t,his h i t i t i , Axll?deol x;l. i tie s e t of all single lions and sum individuals of lions. hiilgnlar c o u n t nouns and count nouris occuring with a numeral (which I will call cuunl nouns" in t he narrow sense) a r e semantically more complex. As I argued for in

hrit'ka ( J987) , they :ire relations bet,ween numbers and en t i t ies build on a measure Suiii~tioti. L,et M! ("na tura l unit") be a function which maps a kind to a measure tunccion %pt?ciJ'i~ for th i s kind, then the meaning of t h e count noun Jion(s) can be r e p r e s e ~ ~ e d a s ~ ~ ~ \ l R ( l e o ' , x ) & NlIdeo ')(x)=r~l . If the number argument is bound by inmi~n i l , a s e.g in two lions,, we ar r ive a t predicates like Ax[R(leo1,x) & NL)(leo1)i x ) = a j , which applies to every sum individual consisting of two lions Note Thai Ihe choice of the form lion or lions is simply a matter of morphological agreement w i t h the determiner, a s I showed in Krifka ( 1 9 8 7 ) .

H o w d o proper names and nominal predica1,es differ syntactically? Proper names normalib { in not require a determiner in English. But note t h a t there a r e cases where a proper name comes with a definite art i d e , like the Sudan, and t h a t t h e core meaning 01 t he definite ar t ic le (reference to a familiar en t i t y ) i s compatible with ihe nu-amng ol a proper name. Common nouns, on the other hand, come in three classes mass nouns like rice can s tand ah an NP without determiner, t he same holds for bare plural nouns l i k e lions, whereas count nouns like l ion(s) require a determiner.

This explains t he possibilit,ies we find for def ini te kind reference in English. A mass noun like gold and a bare plural like lions can be used directly a s a n NP and therefore does not require a determiner if read a s a proper name of a kind. A singular count noun like lion cannot be used a s an N P directly, because i t requires some determiner. The natural choice for proper names i s t h e defini te ar t ic le; and in fact , t h e NP the lion can serve a s a proper name of a kind. A plural count noun like

Page 14: An Outline of Genericity

lions i s morphologically identical t o a bare plural and cannot be distinguished from them. Finally, i t is possible t o use both t h e defini te a r t ic le and pluralization t o refer t o a kind, e.g. in the lions, bu t t h i s sounds a l i t t l e bi t awkward, because e i ther t h e definite ar t ic le or pluralization would be enough t o meet t h e syntac t ic requirements. I t ge ts even worse with t h e combination of def ini te a r t ic le a n d mass noun, e.g. the gold: here t h e noun already i s a n NP and needs no determiner a t all.

33.a) *Lion/The lion/Lions/?The lions i s /a re ext inct . b) Gold/*The gold i s get t ing more expensive.

Up to now, we have only considered kind-referring NPs in subject position. I t i s interest ing t o look a t them when they occur in object position, a s for example with t h e verbs e.xt1nguish and invent .

34.a) The se t t l e r s extinguished the lion/?the lions/'lions. b) The Sumerians invented the wheel/?the wheels/*wheels.

In th i s context , t he definiteness of t h e object NP must be formally marked, if t h e NP is intended t o refer to t h e kind leo. The reason probably i s t h a t in postverbal position, bare NPs tend to be interpreted a s indefinite, so t h a t there must be a formal indicator if a definite interpretat ion i s intended. I will not go in to th i s case of non-compositionality fur ther a t t h i s time,

There a r e interest ing differences between languages in the i r ways of referring to kinds. For example, in German the combination with the definite a r t ic le i s t he preferred means of constructing a proper name referring t o a kind, and can be used even with mass nouns, a s in das Gold.

35.;~) *Loewe/Der Loewe/?Loewen/?Die Loewen ist/sirid ausgestorben. 1)) Gold/Das Gold s te ig t im Preis.

The reason may be t h a t in German t h e combination of definite ar t ic le and proper name is qu i te common, c t . forms like

3 6 ) de r Karl ' t he Charles'

Concerning plural NPs like Loewen: For many speakers , they sound qui te fine a s D- generic isiPs. for others , they sound awkward. There seems t o be some regis ter variation i;oing on; e.g. in biological textbooks, we will find der Loewe 1st :il~sgestorben, whereas in everyday speech, Loewen sind ausgestorben and die Loewen sir ld ausgeslorben a re more common. Curiously enough, definite plurals with n a t i o n a l i t y nouns, like der Turke ' t he Turk ' , underwent a regis ter sh i f t in t he las t decades a n d have now a clearly conservat ive connotation. The explanat ion probably i s t h a t i t is a ra ther conservative way of thinking to consider peoples a s kinds. But I will no1 go into these interest ing regis ter var iat ions fur ther in t h i s ar t ic le .

Let i-is have a look a t French. In French, there i s a requirement. t h a t any noun needs a det,erniiriei- t o be an N P . This explains why i t i s obligatory t,o use t h e definite ar t ic le in t h i s language for D-generic NPs.

37.a) 'Lion/Le lion; 'Lioris/'?Lt?s lions son t e te in ts . b) 'Or!L'or prerid dc la valeur.

N o w , l e t us look at Chinese. In Chinese, any noun is syntact ical ly an NP. Therefore, no determiner is needed at, a l l , a s shown by the following example.

3 8 ) La61iQ juk zh61is ie. t iger vanish CLASS ASP the t iger is ext inct '

Finally, 1 want t o po in t our a n in ic res tmg fact about t he t,ype of def ini te ar t ic le we find in definite generic NPs. In some German dialects , a s well a s in Frisian

Page 15: An Outline of Genericity

(Ebert 19711, t he re a r e two sor t s of definite articles. There i s a long form which i s used for anaphoric reference, and there i s a shor t form t o refer to a n en t i t y which i s par t of t h e shared knowledge of speaker and hearer . This shor t form i s used with proper names (a) a s well a s with definite generic NPs (b) . The following example i s Bavarian:

39.a) Da/*Dea Kare i s kenia. Char les has arr ived ' b) Da/*Dea Schnaps i s daia. 'Schnaps i s expensive' c ) I hab a Bia und a n Schnaps bschdait . Dea/'Da Schnaps war daia .

'I have ordered a beer and a schnaps. The schnaps was expensive'

(c) shows t h a t t h e long form of t h e ar t ic le i s used t o refer anaphorically. 1 think t h a t t h i s example strongly supports t he hypothesis t h a t a generic NP refers t o an en t i ty , and t h a t t h i s en t i t y must be well-established in t h e shared knowledge of speaker and hearer .

We will now t u r n to taxonomic readings. Up t o now. we have singled out two meanings of a noun like lion, one a s a predicate applying to realizations of t h e kind leo, t he other a s a proper name of t he kind leu i tself . There i s a third meaning, namely a s predicate with the extension of t h e sub-species of t he kind leo, e.g. t he Berber l ion, t h e Asian lion, e tc . This count noun meaning even occurs with mass nouns like wine, a s e.g. in three wines. NPs interpreted in t h i s way I call taxonomic; they were identified e.g. by Galmiche (1985) a s formes "pour referer a une sous-espece". I give some exumples for t he taxonomic reading:

40.a) Trie lion is a cat. b ) One lion, namely t h e Asian lion, it; nearly ext inct . r ) Tius l ion ( t h e Asian lion) is nearly ext inct . (1) The rice they grow in East Africa needs l i t t l e water. e ) Tnc l ion and the tiger a r e cats.

^i(: see thul taxonoinic NPs need no t be indefinite, but show the whole range of syntact ic behaviour we observe with count nouns in general.

H o w can W P account for the taxonomic interpretat ion of count nouns in a formal way? keiiienit~ur t h a t lo r the non-taxonomic iri ierpretation, we assumed a mapping NU ("isiatuial i i r~ i t " ) f r o m k i n d s t o measure f i - in~l ions which count individual realizations of t h e kind.-) For the taxonomc interpretat ion, we have to assume another measure function whxvr~ counts subspecies of the kind. Let SS be a function which maps kinds t o such rneahurp f i inct inr is Then we arr ive a t t h e following interpretat ions for t h e count noun 11011 w:

It seems na tura l 1 0 asbume t h a t a subspecies i s a realization of i t s superspecies. Then the two r e a d m e differ just i n their second conjunct: Whereas N U introduces a measure function applying to concrete lions, SS introduces a measure function applying to subspecies of the kind leo

Of course, t he taxonomic readings should be analyzed in terms of taxonomic hierarchies. Basically, these hierarchies can be considered a s la t t ice s t , ructures ( e . g . hay 1971, Pelletier & Schubert. 1983). The two predicational meanings of a noun like lion, t he taxonomic; meaning and the realization meaning, a r e related t o each other; if k ' is a subspecies of k , and s ' i s a specimen of kind k ' , then s ' i s also a specimen of k.

Page 16: An Outline of Genericity

leo ( t h e lion)

t h e Asian lion

subspecies relation - -- -- -- realization relation -----

There a r e some NPs which can only be interpreted a s predicates referring t,o kinds. An example is halogen (cf. Burton-Roberts 1977):

42.a) Chlorine i s a halogen. b) '!There was halogen in t he a i r .

And there a r e constructions which single out t h e taxonomic meaning, namely numerative constructions with special numeratives, e.g. type, or special nominal der ivat ions, e.g. nouns with the suffix -art in German:

43.a) This type of lion is ext inct . b) Diese Loewenart i s t ausgestorben.

The notion of l.axonomic hierarchy can be exploited t o cover t h e type-token ambiguity a s well. For example, this book in (44) can be analyzed a s referring t o a subspecies of t,ho kind book, e.g. to Joyce's Ulysses:

44) This book s o i l s well.

! f 1 y p e - t,oken ainbiguities a r e to be handled a s subspecies-object ambiguities, t hen \\ should b e clt'ar t h a t taxonomic NPs a r e a much more pervasive phenomenon in na tura l l a n g u a g e tiban il was thought before, and sure11 a subject which should be t rea ted K I O I I ' t ho rough ly

'l'here is an i r ~ t e r e s ~ i n g case which supports a n analysis of nouns along these lines, ;ib well as the separat ion of I-genericity and D-genericity. I f t he subspecies meaning of '4 n o u n is a predicate, we should be able to construct I-generic sen tences with I lit:in, t tiat i s , double generic sen tences which a r e I-generic and D-generic a t t h e same time. Curiisider t he following case:

-4!,) A bird flies.

Fhi:; s e i i l e n c e ( ' l t ; ~ - l y h;is two readings, according t o t he interpretat ion of bird a s r e f e r r i n g 1,o individual birds or t o subspecies.

3 . 3 ' I t ie l i , l e t i p ~ i ~ 01' i t--Gener~~ Sentences

b e h a v e analyzed D-genenc NPs a s expressions referring to kinds. In t h i s section, we w i I J t xamnie Itic ont,ology ol kinds, and il,s influence on Lhe meaning of D-generic sen LC-nces

1 x 1 us s t a r t with atialyzing the notion of a kind more thoroughly. Imagine a cage in a zoo with two l i o n s , :ind a fa ther pointing a t them and saying to his children:

46.a) This is t h e lion,

1 think i t is clear that, the lion should be analyzed a s kind-referring NP. Note, for

Page 17: An Outline of Genericity

example, t h a t (46.a) can be continued with t h e generic sentence

46.b) I t l ives in Africa.

where it should clearly be analyzed as referring t o t h e kind leo and a s co-referent with the lion in (46.a). But on t h e o ther hand, t he f a the r pointed a t two concrete animals, which surely do not make up t h e whole kind leo.

How can we solve th i s problem? One way could be to analyze the lion not only a s a n expression referring to a kind, but, a lso a s a predicate applying t o a l l single lions and sums of lions a s well. The D-generic NP the lion could e i ther be analyzed a s ambiguous between those two readings, or a s a n unambiguous predicate with an extension comprising t h e kind leo and any single or multiple realization of t i t . Both assumptions would explain why (46.a) is possible: t h e en t i t y pointfed a t a r e two lions, and therefore in t he extension of the ljun. But i t would be hard t o explain why (46.b) i s a possible continuation, because It clearly refers t o t h e kind leo, and not to some concrete lions

Another, more promising way is to seek un explanation not in language, but, in ontology, i l lat is, not in a wider meaning of D-generic NPs, bu t in t h e ontology of kinds. in t h i s approach, a n NP like the lion should refer t o a kind, leo, bu t the ontology of kinds allows sentences l ike (46.a) to be t rue. The basic ontological difference between objects and kinds can be captured a s follows: For any time t , a n object has a t most one location a t t , whereas a kind may have more t h a n one location a t t. (cf . Langford 1949 and Zemach's (1970) distinction between between "objects" and "types") . A kind, then , i s t o be analyzed a s identical with al l i t s realizations and sums of realizations For example, t h e kind leo i s identical with a n y single lion or any sums of lions. In t h i s way, we can explain why (46.a) i s correct, and why the r i ~ ~ a p t i ~ r ~ r relation belwedi ~t in (46 b ) add the lion i n (46.a) is possible.

of coiirse, t h e oi~iology of objects r i tn i kinds sketched above needs more elaboration. MI nnporlaiit :ispt?ci is t h a t t h e idcii1it.y robit ion has to be relativized Lo different, c r i t ~ r ~ i 0 1 idenl i t ) Otherwise, we could prove t h a t , e.g. , any t w o lions a r e the ; - i : i i i i f , bt;r;iuse t h e y a r e identical to t he kind leo, and ident i ty i s a t r ans i t i ve and 3 o n ~ m i i ~ a t ive relation. Actually, i t i s qui te easy to define the ident i ty relation we m - v d . I t 3 i ils ast,+~iiu) Lwu sor t s of individual en t i t i es , kinds and objects, and i :it$lfaon's 1\~1o-pi:ic'e relation R. Then we can define a relation EQU a s follows: V A . y ~ I + " ( , J ~ ~ ( A , ~ ) w [ x = y i v R(x,y ) I Here, "=" should denote ordinary equal i ty .

$ 1 ( f i l s point,, 11 1b interesting t,o consider rhe relation between kinds and the t %tc~iisioii of rioiiiin:tl predicat,es again. In section (3 . 1 ), I showed how the meaning of nominal predicams c a n be construed froin t h e representation of a kind. If we assume the CQI I r c i a t i o i~ as basic, I h e n we L-KII gel rid of t h e R relation; for example, we could it.*~re.stirit t h e pi*t-dic~te g o l d as AxILQIJ(Au',x) & -ix=Autl. But we s t i l l have t o m a k e ~ p h o t t h a i t h e predicate gold is related to some criterion of ident i ty which a l l o w s u s to s ay , t o t - oxainple, that the y u a ~ i l ~ i t y of s tuff I found yes te rday and the qaarititv or stuff i n my pocket is the saine gold. It is well known t h a t cr i ter ia of identily a n dependent on the kind, for exanipIt1, t he s tuff 1 found and the s tuff in rny pocket may f c i ~ I he same quantAty of gold, but different rings. To capture th i s depeni-lenc"), I propose a relation I l l which m a p s a kind to en t i t i es with t h e cr i ter ia of i d f i i t i p sperifio for t h e kind Then we ger representat ions like t h e following ( H I P S :

There a r e ways t o handle en t i t i es i ~ - ~ i t t i different cr i ter ia of ident i ty; cf. for example Gupta 0 9 8 0 ) H:isic:~lly, Ihr variable x in (46') can be considered a s a sort, i f predicate varirtble applying to more basic en t i t i es , possibly temporal s tages in t h e sense of Carison ( 1 9771 Then the Furmula lD(Aul,x) says t h a t x i s a n equivalence class i u r t h e kind An, e . g . tha t all s tages a for which x ( s ) holds a r e s t ages of t he

Page 18: An Outline of Genericity

same quant i ty of gold. Note t h a t t h e notion of a s tage , i n cont ras t with Carlson's work, would play a modest role i n such a theory: it jus t helps to give a n formalization of cr i ter ia of ident i ty .

Let u s come back t o D-generic sentences. If t h e ontology of kinds given above i s correct, then t h e following two sentences should both be appropriate to describe a s i tua t ion which consis ts of a man wearing a h a t going across a s t r e e t in to a pub. They exemplify just different modes of speaking, which I will call object-oriented and kind-oriented, respectively.

47.a) A man wearing a ha t went across a s t r e e t into a pub. b) Man wearing the h a t went across t he s t r e e t into t h e pub.

But clearly, t h e kind--oriented mode of speaking in (47.b) sounds r a the r odd. The reason why we normally employ the object-oriented mode is t h a t i t usually i s more informative, a s i t provides u s with quant i ta t ive information and with cr i ter ia of ident i ty which help us to keep track of t he same object by anaphoric devices. For example, in (47.a) i t i s expressed t h a t oiie man, one ha t , one s t r e e t and one pub a re involved in the reported even t , and i t i s possible to refer to t he same objects la te r , e.g. in He found out that i t was closed. I n (47.b). however, i t i s unclear how many men, ha t s , s t r e e t s and pubs a r e involved, and i t would be impossible to refer explicitly to t h e same objects in following sentences. This explains why we mostly speak in t h e object-oriented mode in t he f i r s t place.

B u t when do we employ the kind-oriented mode of speaking? I will discuss t h e cases I a m aware of, bu t there may be some more.

48.a) Man s e t foot. on the Moon in 1969. (avantgarde interpretat ion) b) be quiet - i h e lion i s roaming about! ( representa t ive object) c ) Frenchmen c;i t horsemeat. (distinguishing property)

The l"r~:nchman e a t s horsemeat. (1) (irrinans bought, 83000 BMWs l a s t year . (collective property)

The German bought 83000 BMWs l a s t year . t l ) The (iodo is extinci (kind predicate) 1 ) fhn lioit is ferocious. (character is t ic property)

a ) I f some exceptional objec~, has a property, we can employ the kind-oriented mode lietaaorie the property is fe l t l o be relevant not only to t h e object,, but, a lso to t he k i n d i h i b object belongs to. 1 call this t h e avantgarde interpretat ion. Note t ha t not ; m y old exceptional properly object gives r ise t o th i s interpretat ion, because i t is n u t i icc~ssar i l ) felt Lo be relevant t o t h e kind. This can be seen by t h e following ~ J X ~ I ~ ~ P ~ ( ~ S :

49.a) Man learripci Lo so lve cubic equat iuns in t he 16th centjury. b) ?Man jumped over 8.90 meters in 1968. v ) 'Man a l e 128 pretzels in one hour in 1976.

b) It' t h e object is only interest ing a s a representat ive of a kind, then t h e kind- oriented mode is employed a s well. 1 f a l l th i s representa t ive object interpretat ion. This is the interpretat ion we have to assume for our ini t ia l example (46.a).

c ) Another ease w h ~ r e t he kind-oriented mode of speaking i s called for is when i t i s a distinguishing property, that. is, a property t h a t holds of some realizations of the kind and t h u s distinguishes t h i s kind from other , closely real ted kinds. Note t h a t this reading s h o u l d not be t iea ted in terms of I-genericity, since no defaul t ~ u a n t i f i c a t i o n i s involved hero ( e y, , there is nu default rule that a Frenchman ea t s iiorsemeal). But h o w can we show that 1 ) - g e n e r ~ c sentences with distinguishing propert.> interpretat ion a re iiifferent from 1- generic sentences? Note that, with I-generic readings bare plurals and iridefinile s ingulars can be exchanged salva veritate, whereas D-generic NP.s in Ihe distinguishing property interpretat ion cannot be exchanged salva v m t a t e with inch-finite siiigular Nl's.

Page 19: An Outline of Genericity

50.a) Lions a r e ferocious. b) A lion is ferocious. (meaning of a = meaning of b)

51 .a) Frenchmen e a t horsemeat. b) A Frenchman e a t s horsemeat. (meaning of b s t ronger t han a )

Note t h a t t he following famous example (cf. Carlson 1977) can be explained similarly a s D-generic sentence with distinguishing property interpretat ion, a s ( a ,b ) in t h e most plausible interpretat ion differ in meaning from (c):

52.a) Dutchmen a re good sailors. b) The Dutchman i s a good sailor. c) A Dutchman i s a good sailor. (stronger t han a and b)

d) The kind-oriented mode is allowed when a predicate applies to t he sum of al l existing realizations of t h e kind. I call t h i s t h e collective property interpre- ta t ion ; i t was alluded t o in Carlson's example Linguists have 30000 books in print.

One could argue by these examples that the proper way to analyze kinds is the mereological one, t h a t i s , to analyze a kind a s t he sum of i t s realizations. But I Ohink we should carefully distinguish those en t i t i es , a s not every predicate which holds for t he sum of t he realizations of a kind holds for t h e kind i tself , and vice versa For example, i t may be t rue Ltiat. t h e sum of t he existing lions have a weight of 100000 tons, but a generic sentence like

59) 'The lion has (or: Lions have) a weight of 100000 tons

is s imp ly nonsense. It remains t o be invest igated under which conditions the +ol lec t ive properly int,erpret,at ion is available.

4 % ) Kind predicates r a n only be applied t o kinds in t.he first place. But note t h a t ilif-iy an' re la iud l o properties of realizations of t h e kind. For example, in order to , t l ow t h a t t he dodo is ext inct , one has to show t h a t there have been realizations of nil:; hmd it1 the' pa;>{, and t h a t there a r e no present realizations of t h i s kind now 1 itis r'fcdu(-tion nt' kind predicates to predicates about objects can be handled by nieanitig postui,,Lt-s lor kind predicates.

I ) b ' ina l l~ , if t h e realizations of a kind have a character is t ic property, then th i s ~ t o p ( x t y (-an be "projected" t u t he kind i tself . I call t h i s character is t ic property n~ierpre ta t ior i In these cases , both D-genericity and I-genericity a r e involved. For tix:iiftpiti, ij-.genonr bent ences witti t h e character is t ic properly interpretat ion can have ail t h e over t qiiuniiliers we have found a s a l te rna t ives to t h e I-generic ( onst rur l ion:

f)"t.a) A l i o n always mostly/usually~aelocmi/r~ever has a mane. b ) The lion always/rnostly/ usually/st.-ldoin/never has a mane

h u r e l y , w t ? h a v e intuitively the same kind of default rule in the lion is ferocious a s in a 11u1i is ferci-iuu^ Therefore , t h i s sentence should be analyzed as a sentence which contains both the i)l';[-'AIJLT quant if ier and a D-generic NP.

How can we capture th i s sor t of double genenci ty formally? According t o t h e ontology of kinds sketxhed above, the predicate a lion and the kind-referring NP the lion a r e semantically ra ther similar. Of course, a lion denotes a s e t , and the lion denotes a kind individual. R u t every element in t he extension of a lion i s equal to t h e kind individual denoted by t h e lion (according t o t he extended equal i ty relation introduced above) . This suggests the following formalization:

55) The lion hat- t i inane. i)E~~~~~~[x~.y~[x=leo'],[mane'ty), hasf(x,y)D

Page 20: An Outline of Genericity

Although t h e variable x in t h e an tecedent pa r t i s bound t o a specific individual, t h e kind leo, there a r e many possible ins tan t ia t ions for i t - namely any single lion and any sum of lions. That is , t h e antecedent pa r t has t he same meaning as t h e expression lions'(x), assuming t h a t lions' applies t o single and multiple lions. But t hen the formalization of (55) i s qu i te plausible. The only refinement necessary i s t h e inclusion of a dis t r ibut ivi ty operator in t he consequent, because x can be ins tan t ia ted by a sum of lions as well. As it. i s often t h e case, t h i s dis t r ibut ivi ty operator i s implicit, bu t i t can be made explicit , a s in German Der Lowe hat je vjer Re imhne ' t he lion has four fangs each' .

In t he examples considerd so fa r , we mostly analyzed kind-referring NPs in subject position. But they can occur in other syntac t ic positions a s well, e.g. in object position. The range of meanings, however, seems t o be limited in these cases . At leas t the following readings a re clearly available:

56.a) The s e t t l e r s ext inguished the dodo. (kind p red ica t e s ) The Sumerians invented t h e wheel.

b) In Tanzania, they filmed the lion. ( r e p r e s e n t a t i v e o b j e c t ) c ) Dogs chase cats.

The dog chases the cat. ( c h a r a c t e r i s t i c p red ica t e )

The character is t ic predicate reading typically shows up when the subject i s also a kind-referring NP. Normally, t h e object NP should be clearly marked a s def ini te when reference to a single kind i s intended, but i t could be t h a t cats in (c) can be interpreted a s referring t o a kind a s well. I will not go in to these problems here.

My t reatment of the interpretat ions of sentences containing D-generic NPs (or. eyuivalent ly, circuri~stances t o l icense the kind-oriented mode of speaking) differs Irom approaches like the ones of Bacon (1973) and Heyer (1985), who assume different interpretat ions of the D-generic NP il.self to handle t he interpretat ions of t h e sentences For example, according t u Heyer t he D-generic NP the lion can refer e i ther 1 0 The k i n d Itw ( in the kind predicate and avantgarde interpretat ion, h i s "absolute generic retererioe"), or to a typical representat ive of t h i s kind (in t,he character- is1 ic property interpretal ion, Iiis "personal generic reference"). But pushing t h e iiflers-in,e to t h e referent of the D-generic NP seems to be incorrect, a s i t i s t he v ~ r t i : i l predicate which is decisive for the interpretat ion of t h e sentence. For c x a i n p l ~ . one a n d t h e same NP can serve both roles in sentences with conjoined verbal pvçi ih- : i l i ;h as i n nit,' dodo lived on Mauritius and became extinct in the 18th century

i t in i f l i t h e wonliwile to look a t genericity again a f t e r we worked through a l l t h i s re-claissifieation. Basically, we found t h a t D-genericity can be analyzed a s reference t.o k i n d s , whereas I-generic ' i t .~ can be analyzed a s default quantification.

I t was not qui t ( : easy to separa te these two notions. The reason i s t h a t there i s considerable overlap in f u n n a s w e l l a s i n meaning. Consider first, t h e overlap in form. Three of t he five types of D-generic NPs can be analyzed a s indefinite NPs a s well (namely bare mass ternis, bare plurals and indefinite taxonomic NPs) and can therefore be interpreted a s I-generic NPs in appropriate contexts .

Formal ovbrlap of D-generic NPs and I - g e n e r i c i t y 57 d e f i n i t e s ingu la r NPs X k ind- refer r ing NPs

d e f i n i t e p l u r a l NPs X bare s i n g u l a r s X X bare p l u r a l s X X i n d e f i n i t e taxonomic NPs X X i n d e f i n i t e NPs o the r NPs with i n d e f i n i t e de te iminers X (poss ib ly I -gener ic )

Page 21: An Outline of Genericity

Concerning t h e semantic overlap, we have seen t h a t a t l ea s t one qui te common interpretat ion of D-generic sentences, t h e character is t ic predicate interpretat ion, must be analyzed a s habi tual interpretat ion, similarly t o I-generic sentences.

Semantic overlap of D-generic NPs and default quantification 58) avantgarde X

representative object X interpretations of sentences distinguishing property X with kind-referring N P s collective property X kind predicate X characteristic property X X I-generic sentences X default quantification habitual sentences X

Given th i s overlap in form and meaning, I think i t comes a s no surprise that. a ra ther iinil'orni notion of "genericiLyt' has been favoured for such a long time. But a s I have argued, I think we should abandon th i s notion in favour of two o ther , re lated notions, tiaiiiely kind reference a n d defaul t quantification.

in this final chapter , I want t o extend the analysis of 1-genericity and habi tua t iv i ty given above in various ways.

l 'he relation of any and the so-called indefinite generic ar t ic le is a much-discussed lopic (set? e . g . liurtofi-Roberts 1976). In t h i s section, I will show t h a t i he t i ibtr ihut i~ri 01' m y earl be explained in our theory of I-genericity.

-\s I h a v e munlioried a b o v e , there i s a n important distinction between 1,-generic s(:nttinces a n d sentences including nominal quantification, noted by Dahl (1975) and, i f i n n * ~ ~ x p i i r i l l y , by Croft (1986). In cont ras t to nominal quant if icat ion, I-genericity ,-..iiinul be reslrici ed LO a. given set of individuals:

b;)) I ' I i t - r t ? * t 2 r à lions and tigers 111 I he circus ring. a) Every l i o n ~ a l l l i o n s / ~ a c h lion/inosr lions/many lions/no lion had a inane. h) ions liail rriano

In (a ) , the rarige o f nidividualb qi.innUficd over is restr ic ted t o t he contextually salieiii ones, wliercas an I-generic sentence like ( b ) cannot be restr ic ted in th i s w i i y . There is, however, ri qi-iautifier wh ich cannot be restricted e i ther , namely any:

6 0 ) A I I ~ l ion has a inane

11 is evideni t ha1 ( 0 0 ) has 1 ha chararu~rihl i t : property of I-generic sentences; for axample, 11 need5 a s t a t i ve , non-accidental predicate (cf. * A n y lion is roaring), and HIP rfinr.t i ) l ' quan t i l na t ion (,aniiol. be restricted to the cont t~xtunl ly sa l ien t individual;-; a ' f . \lt-ndler 1967 and Perlmutter 1970 for t he genericity of sentences like 60).

i i u w can w e account for t h e fac t t h a t an N P l i k e any lion can be the subject of a n 1 - generic s e n t e n w ; O u r analysis of 1-gerii-ri~ i t \ requires t h a t i t is basically an indefinite NP insofar a s it h a s no quantiIicaliona1 force of i t s own. Traditional w i s d o m s a y s that 11 is a negat ive polarity item, and therefore occurs in t h e scope of downward- entai l ing operators according t o t he theory of Ladusaw (1979). This insight can be combined with our theory of 1-genericity a s default quantification if we can

Page 22: An Outline of Genericity

show t h a t t h e antecedent of t h e defaul t quant if ier i s a downward-entailing context.

Let u s begin with t h e f i r s t claim, t h a t any lion i s semantically similar to a n indefinite NP. I cannot go in to the involved question concerning t h e analysis of any here, bu t simply note t h a t Ladusaw (1979) offers arguments for a n analysis of any a s a n ex is ten t ia l quant if ier (specifically, a n exis tent ial quant if ier with narrow scope), and aga ins t another common analysis , namely a s a universal quant ifer with wide scope. In Ladusaw's analysis , a n NP like any lion resembles a n ordinary indefinite NP like a lion in t h e Generalized Quantif iers framework. This resemblance should carry over t o a n analysis of indefinite NPs in t he Kamp/Heim framework, where indefinite NPs have no quantificational force of the i r own. The narrow scope condition can be recast a s a condition saying t h a t t he variable of a n any-NP must not be bound by t h e exis tent ial quant if ier ranging over t he whole t ex t .

The second claim, t h a t t he antecedent of t he defaul t quant if ier is downward- entailing, clearly i s incorrect in t he general case. Although t h e an tecedent may be s t rengthened in many cases (cf. 61) , there a r e some cases where we ar r ive a t wrong conclusions (cf. 62) :

61) A lion has four legs. A male lion i s a lion. - > A male lion has four legs.

62) A lion has four legs. A three-legged lion i s a lion. /-> A three-legged lion has four legs.

The reason why we cannot draw the conclusion in (62) i s t h a t t h i s specific strengthening of the antecedent res t r ic t s i t to cases where the defaul t rule i s not applicable anymore. In the example a t hand, t h e consequent, t h a t x has four legs, is no1 ronsisU?n t w i t h t he positive iiiforinat.ion given in t he a r ~ t e c e d ~ r i t , which says that ,

is three-legged (We meet similar problems with conditional sen tences in general, 1 H e i i i i 1987) l'tiarefore w e have Lu exclude those s t rengthenings of t he antecedent v+tiicli bear inlluence on the consistency of t he consequent. But if we keep th is restrieliori in mind, the antecedent o f t h e DEFAULT quant if ier i s in fac t downward ~Lail i i ig . And th i s explains why it l icenses any-NPs,

I ' hu (.loi~/nward-enlailrnent property explains furthermore why sentences with indefinite geiienc NPs conjoined by o r can have a n interpretat ion equivalent t o single generic s e n 1 erices conjoined by and:

62') A lion or a horse have a mane. < - > A l ion has a inane and a horse has a mane

According t o our theory, these sentences have roughly the following interpretat ion:

This infererice is s tandard i f we replace DEFAULT with universal quant if icat ion. But i t i s clear That il shou ld hold with DEFAULT a s well: If a n x which i s a lion or a horse has a mane by defaul t , than th i s defaul t holds for a n x which i s a lion and for an x which i s K horse separately us well.

Note t h a t (62 ' ) can t iabe an interpretat ion w i t h wide scope o r a s well. It' we look a t sentences with NPs conjoined by and, we only find th i s wide scope reading acceptable (cf , A lion and a horse hai,tb a mane). The reason for t h i s i s clear: otherwise we would end up witli a n ante(-chdont, like [l ion'(x) & horse1(x)1, which does not hold for any x. But if arid i s interpretxd as construing a sum individual, t hen we end up with a n acceptable sentence, a s e g. in A lion and a tiger hate each other.

Page 23: An Outline of Genericity

Let u s come back t o any. There is an Interest ing meaning difference between I-generic sentences with any-NPs and ordinary indefinite NPs:

63.a) A lion h a s a mane. b) Any lion has a mane.

(63.b) i s interpreted a s more s t r ic t ly t han ( a ) ; in fac t , (a) can be considered a s t rue , whereas (b) i s clearly false . This difference can be explained by the as- sumption of Vendler (1967) t h a t any-NPs convey the meaning t h a t nothing hinges upon the part icular choice of a referent,. The DEFAULT quant if ier lacks th i s property; in fact , whether t h e consequent applies or not often depends on the selected referent,. Therefore sentences like (63.a) do allow exceptions, whereas sen tences like (63.b) do not .

Burton-Roberts (1976) notes some differences in acceptabili ty between a-NPs and any- NPs in generic sentences; one o f her examples is A beaver / 'Any beaver is an anlptiibious rodent. The reason why the any-sentence sounds odd seems to be t h a t t h e predicate expresses a necessary, nearly definitional property which holds for every beaver. But. then t h e "strengthening" of a beaver t o any beaver does not make much sense , because the defaul t rule held without exception in t he f i r s t place. This explanation can be made more precise if we assume t h a t a sen tence with a n any-NP is more complex than the corresponding sentence with an a-NP, and t h a t t h e simpler expression i s used if t h e more complex one does not differ from i t in meaning.

Nunberg & Pan (1975) t ry t o capture t h e difference between a-NPs and my-Ws by assuming: t h a t t h e extension of t he a-NP i s restr ic ted to individuals which a r e typical representat ives , whereas t h e any-NP is t o be understood more widely, includir~g all the u n t y p i c a l cases as weli But then we cannot explain why sentences l i k e ' tins is any l ion , pointing t o a specific hon which possibly i s not very i vpitbai For its specie';,, a re bad.

i'tiure 1s a ('lass 01' predicates, usually considered a s kind predicates, which can be applied to inde l imte NPs and to kind-referring NPs a s well. Examples a r e be rare, be romition, he widespread.

hF> : i ) T h p lion 15 ra re . b ) The lions a rc rare . c ) 4 lion ( w i t t i three legs) is ra re . d ) Lioti:, ( w i t h three legs) a r e ra re . o ) Gold (wh ich is hammered f la t ) is rare . t ) (me cat ( i i i i ~ ~ ~ l y t he lion) is ra re

Obviously, i h i s distribution pa t te rn cu t s across t he pa t te rns we have observed in sertiori ( 1 . 2 ) , and w v have to account for i t in some way.

I think t h a t be ran- s h o u l d not be analyzed ah a kind predicate, like be extinct, b e c a u s e i l can b e combined with a s ingular indefinite NP (cf. 59.c), and i t need not be applied to a well-established kind O n t h e other hand, t he examples above cannot, be analyzed a s 1--generic sentences e i ther . They do not. express some defaul t rule which holds for individual l ions, and they a re not related t,o sen tences with overt adverbial quantificatioii , cf .

6 6 ) * A lion ( w i l t 1 three legs) is a lways /~nos t ly /of ten /us~~al ly /never ra re .

They do not allow ariy-NPs, cf. *Any lion is rare. And finally, a predicate l ike be rare cannot apply t o an individual l ion in t he first place.

Page 24: An Outline of Genericity

I t i s possible, however, t o analyze predicates l ike be rare as semantically complex applying t o t h e denotation of nominal predicates, i.e.. indefinite NPs. A possible analysis along these lines i s t h e following one. Note t h a t I do no t claim t h a t t he specific analysis of rare i s t h e ultimate one; I only claim t h a t we should seek for some analysis of t h i s sor t .

67) A lion with three legs i s rare . raren(Ax[lion'(x) & has-three-legsn(x) 1) with t h e following meaning postulate for rare': raren(P) => there a r e few locations 1 for which i t holds

t h a t there i s an x a t 1 with P(x).

Let u s call predicates like be rare set predicates, because they apply not t o a kind, bu t to a s e t of objects.

But how could we handle sentences like (65.a) , which clearly a r e predication t.o a kind? According l o t he ontology of kinds sketched in section (5.3) above, there is a close resemblance beifween the denotat ions of t h e predicates a lion or lions and the ontological s t ruc tu re of t he referent of t he kind-denoting NP the lion. All we need is t o represent t h e kind by i t s singleton s e t , triggered by some type-shifting rule.

6 8 ) The lion is rare . rare'('/.xlx=leo'l)

According t o t h e

b 9 ) there a r e an x a t 1

We also have s e t the verb love i n

meaning postulate in (671, t h i s amounts to

few locations 1 for which i t holds t h a t there i s wi1.h x=leo'

predicates which require a se t - type object NP. A case in point is a special reading. The verb Jove can e i ther be a relation between

t w o iri(Jivnlu;ilh (cf 70.a,b), where t h e object can denote a kind a s well (cf. c ) . Or I Â ¥ a denote a relation between a n individual and a s e t , meaning t h a t t he individual l o v e s t o have a n eltr?tnent of t h e set, (cf. d ) .

7 u . a ~ John loves Mary. t ) ) John lovcs a woman. c ) J o h n loves t h e IBM PC. d ) J o h n loves a hot shower . (= John loves t o have a hot shower).

i n t h o classic l'oniial t reatment of generic it,^ developed by Carlson ( 1 977), every bare plural NP is analyzed a s referring to a kind. As generic NPs can be qui te complex ( Ã g i f o i~s with three^ legs) , and a s there seems to be no principled limit for the i r co~~ip l t . -~ . i iy . one has to assume qui te a lo t of individuals. Nothing less t han a d i s t ~ ~ I individual for every possible noun denotation would do. Furthermore, ( ,arlsonts analysis can be extended t o cover sentences l ike t h e following one, which can t ~ t l .irirflyzed as prtldic-ations t,o t he kind of jogging (cf. Chierchia 1982).

71 .a) Jogging is healiriy. b) To jog is heait,hy.

' I h i s leads u s to t h i n k about representing every possible predicate, nominal ones a s well a s verbal ones , by an individual in the basic domain of the model s t ruc ture .

01' course, this approach presents problem?? u we assume standard models. Chierchia (1982) and Turner (1983) showed possible w a h s to overcome them by drawing on a s y n t a c t i ~ ~ i l r e s l , r ~ v l ed represtlrii nt ion langunge developed by Cocc-hiarella a n d using

Page 25: An Outline of Genericity

Scott Domains a s model s t ruc tures , respectively

In t h e analysis of genericity developed here, however, we actual ly need only "few" individuals t o represent kinds. This i s because we need them just t o represent t h e denotation of D-generic NPs, and as we have seen, we cannot construct D-generic NPs from any predicate whatsoever, bu t only from predicates which a r e related t o a well- established kind (cf. 67.a,b). A l l t he cases which seem to show t h a t we need "many" individuals can be t rea ted e i ther a s I-generic sentences (cf. c) or a s sen tences involving s e t predication (cf. d l , and we do not need kind individuals t o handle them.

72.a) *The lion with three legs i s ext inct . b) 'Lions with three legs a r e ext inct . b) Lions with t8hree legs a r e severely handicapped. c) Lions with three legs a r e ra re .

If t h ~ t reatment developed here i s basically correct and can be extended t o cover sentences with subjects l ike jogging, to jog, then we need not introduce complex model s t ruc tures in order t o model t h e semantics of genericity. In t h e case of common nouns, I have sketched a theory which assumes kinds to be primitive, and nominal predicates to be derivat ive constructions. Take the simple example gold. We have a kind Au and a derived nominal predicate ^.xlR!Au8,x)] or Ax[EQU(Au8,x) & ID(Au8,x)] (cf. section 3.3) applying to realizations of t h i s kind, t h a t is , quant i t ies of gold. The complex predicate gold which is hammered flat can be represented by t h e appli- cation of a predicate modfier like ^.P^.x[P(x) & hammered-flat8(x)l t o t h e nominal predicate, t h a t is , by ^.x[R(Au8,x) & hammered-flatS(x)j or Xx[EQIJ(AuS,x) & ID(Au8,x) & hammered-flat8(x)1. But t he re i s no need for a kind individual which has a s i t s realizations just t h e quant i t ies of gold which a re hammered f la t , a possibility t h e theories of Carlsou and Chierchia do allow for.

11 hctxns possible to carry o v e r t h i s approach to verb nieanings a s well. Take jogging i-os an exainplt?. We (-an assume a kind jog which has events of jogging a s i t s I t a1 ite.a tioris. Thori w t? can construct an event p r e d i c a ~ e , ?Le[R(jog8,e)] The complex p r ~ ( i ~ c h . t u jog l i d t :t wilr l hen can be represented a s Ae[R(jogS,e) & half-a-inile8(e)j, mid there need rit!l b~ any k i n d which curretaponds to t h i s s e t .

( I ! ' course, language users a r e ra ther f ree to assume even ra ther s t range kinds, and therefore one could argue t h a t i t is wise t o presuppose a complex model s t ruc ture in I n - f i r s t place But t h i s would be agairist the spir i t of much of t h e r e s t of lingu- isl.ichi, e.g in s y n t a ~ , where we try to rest'rict the power our theories as f a r a s i ) O i i S l i ~ ~ t ' And i t viouid blur an important seniantic distinction, namely t h a t of well- ~s tab l i s t ie i i kinds arid predicates which a rc arbi t rar i ly construed.

llrre, 1 j u s l waul t o mention t h a t it i s no[ sufficient to assume just kinds a s individual corrc'spor~dents of some riouns. For example, we have not only t h e kind homo s a p i e n s c o r r e s p ~ i ~ d i n g Lo the count noun tiifin, but also a n individual which we denote wi th n i n n k i n d T h a t ihere a r e subt le differences between those individuals can be seen in the following example:

7'2' a ) itlan s e t tool o n t he Moon in 1969 b ) ?Mankind sel loot o n t h e Moun i n 19t'i9.

The referent of n i a r / / i i ~ ~ d probably is t h e sum individual of all men, and (b) sounds odd because o n l y a very small part of th i s individual actual ly s e t foot on t h e Moon in 1969.

Another type of quasi-generic individual can be assumed for N l * s which contain adjectives like typical or f u m e , a s in

72'I.a) The typical German spends Christinas with his family. b) The average Gerniari ea rns DM 32000 a year .

Page 26: An Outline of Genericity

These sen tences can be explained most easi ly if we assume t h a t the i r subjects refer to hypothetical individuals which have exact ly those properties which a r e supposed to be typical or common for t h e individuals denoted by t h e head noun. Then any a t t r ibu t ion of a property t o th i s individual en ta i l s an a t t r ibu t ion of t h i s property t o elements i n t he extension of t h e nominal predicate.

The meaning of sen tences l ike (72'') i s close t.o D-generic sentences in t he character is t ic property interpretat ion. This may be a reason why cer tain authors , for example Heyer (19851, reconstruct definite generic NPs in t h i s interpretat ion a s referring to a typical representat ive of a kind. But we have presented arguments why a uniform analysis of D-generic NPs for a l l interpretat ions is necessary. Furthermore, note t h a t we can identify a concrete individual with t h e typical representa t ive of a kind, as e.g. i n Fritz is the typical German, whereas t he identification with the kind itself is licensed only in very special contexts , viz. ??Fritz is the German (referring with the German to t h e kind).

A historical remark may be in order at th i s place. Heyer (1987) has shown Lhat generic s en ten res were a central lopic i n t h e tradition of ontological Realism; they have been considered even a s more basic t han sentences involving concrete ent i t ies . With t h e r ise of Co~~ceptualism/Nominalism, however, generic sen tences were considered to be a secondary phenomenon, giving evidence of t h e inconsistencies of na tu ra l language. For example, Frege (1892) only al ludes to generic sen tences l ike D a s Pferd 1st ein vierbejriiges Tier ' t he horse i s a for-legged animal" and Der Turke belagerte Wien ' t he Turk besieged Vienna', but fai ls t o give a uniform treatment , and Russell and Quine only consider quantification over objects. Modern theories, a s Carlson (1977), Chierchia (1982) and Heyer's own theory can be considered a s a complete return to ontological Realism, insofar a s they assume kinds and properties in t h e domain of individuals. In t he theory of generieity developed here, we distinguish tmtweon two so r t s of geriericity. A Realist theory is called fur only for D-generic W s , whereas 1-generic sentences can be reconstructed in the Nominalist way a s a special sor t of quantification over individuals. This i s not simply a compromise, but is suggest,ed by a careful examination of the linguistic phenomena.

Although i t h a s never been an important topic in t he discussion, i t i s obvious t h a t [-generic sent,enres differ from noii-generic sentences in accentuat ional fea tures :

73.21) A tiori ! is a ferocious 'beast, b ) A ' l i o n is id the cage,

A i i l - ~ ~ , t : ~ ~ . ' r i ~ ~ sen ten re l ike (a) has two par t s , which may be identified a s theme and rheme There i s an optional pause between them, the rheme bears t h e main s t ress , and the theirio bears secondary s t r e s s . Sentences with normal indefinite subjects like (b) d o n o t have a n optional pause, t h e subject bears the main s t ress . and there i s a i t2iid~sic.v to use the therc-constructiori. They can be considered a s wholly rhematic. That )-generic NPs hire thematic becomes very clear in languageas like French which h a v k :1 special construction for themalit: indefinite const i tuents (a ) , whereas t he common construction for rhematic indefinite const i tuents ( b ) cannot be interpreted as gt'nenc:

74.a) Des garcons. (:a rm pleurent pas , 'Boys do not cry.' ti) I1 y a des garcons q u i pleureril.

Note lliat, the sentence pa t te rn of I-generic sentences i s common to a l l sentences with quantificational s t ruc ture :

Page 27: An Outline of Genericity

75.a) Every lion ! i s a ferocious 'beast . b) If Pedro h a s a donkey I h e 'beats it.

This observation can be explained if we associate the Kamp/Heim notion of quantifiers with t h e themehheme distinction developed in t h e Prague school. This can be done qui te natural ly: The antecedent of a quant if icat ional operator forms t h e background against which t h e consequent i s eva lua ted ; and t h e relation of theme to rheme can be thought of in exact ly t h e same way. Now. i t i s well-known t h a t t he theme/rheme- s t ructur ing bears influence on intonation. Basically, t h e rheme contains t h e main s t ress , and theme and rheme a r e separated by a n optional pause. But t h i s describes exactly our findings concerning t h e accentual pa t te rns of quantificational sentences.

Of course, our assumption only holds for sen tences which do not have a narrow (contrast ive) focus. For example, i t i s possible to u t te r t h e following sentences with s t r e s s on lion, without changing the interpretat ion, bu t implying soma element contrast ing with lion.

76.a) V!very 'lion i s a fercocious beast,. b) A 'lion i s a f'erocious beasi,.

To handle these cases, we have to assume a second layer of intonat ional rules which governs phenomena like contrast ive focus. I will not go into these complications here, but, simply res t r ic t my examples to non-contrastive interpretat ions.

Let us examine some more examples to s ee whether t he observed relation between intonat,ion and quant if ier s t ruc ture can be maintained.

W P came across some instances where i t was convenient t o assume t h a t t h e antecedent par t o t t h e q u : t . n l i f ~ e r gets enriched by some more material , e.g. by (atemporal or temporal ) when-( laiises As the when-clause is part of t,he antecedent , i t should be iirist roused I t J.S exactly th i s t h a t w e find:

7 7 a ) L.iuns a re 'dangerous i when t h e y a re hungry. l)Kb'Al~L'l 'l i , t ; ~ , t l ( l l i o n s ' ( x ) , hungry'(x,t~l,ldangerous'(x,t ) ] I

h; 1,ion.s a r c 8 'dangerous 1 when itlev h a v e red eyes. i ) K i % " \ l i l ,T~x ,x l~ l ions ' i x ) . have-red-eyes'(x)],[dangerous'(x)])

-1 second e ~ a r n p l e which our analysis accounts for a r e t h e ambiguities discussed by Carlbon (1986) . Carlson observed t h a t sentences like flowers grow ou t behind the old sfsf d h a v p t w o ineannigs. T h e s e meanings can be rendered easily in our analysis , and I i i r i nt~riii.) r t i , I 11c.~ a r r clearly favored with I he s t r e s s pa t te rns predicted by our ;isboi i h t i o i i o f quant if ier s t ruc ture to theme/rheme s t ruc ture .

,"rt a ) Flowers i grow out, behind t h e old 'shed. PRFA~JI,TIX,~ , - x , t ~ ~ ~ f l o w e r s ' ( x ) , l i ( t I , [grow-out-behind-the-old-shedf(x, t ) l )

h ) 'Plowers pro& out 1 behind the old shed. l )h :KA lJLT1 i, A I ( [ behind-the-old-shed'(t),P(t)],lflowers'(x), grow-outt(x,t)])

(-,rtrlson f u r t h e r n i u i ~ observed t h a t with simple habi tuals , there i s only one reading ( t h e fn'sl one) . This can be explained with a plausible additional principle, namely t h a t t h e antecedent must fcu- specified by at, l eas t one lexical item (otherwise, t he sentence would h a v e rather weak t ru th conditions, e.g. that. there a r e flowers growing somewher43 by f l ~ f i i i 1 1 1 )

79) blowers grow i t ) DEI 'AIJ~.T[X, l ; A , r j([flowers'(x), P(l,)l , I ~ ~ o w ' ( x , ~ ) ] ) b) *~)Eb'~ \ IJIArl ' [x ,x , t l ( l l i ( t ) j , Iflowers'Cx), grow'; k , i ) I )

( n o t allowed; antecedent must be explicitly specified)

Carlson showed i n t i ih I986 paper t h a t gi?r~ericity/habiI,ualit,y i s a relational phenomenon, bu t did not, give a fortnal treatment.. In my analysis , habi tual sentences

Page 28: An Outline of Genericity

get a formally relat ional analysis by t h e DEFAULT-quantifier.

As a third example, I want t o discuss a distinction which was made by Lawler (1972, 1973) and Dahl (1975). They pointed ou t t h a t examples l ike t h e following can have two readings, which they named "existential" and "universal".

80) John drinks beer.

As a universal habi tua l sentence, (80) means t h a t beer i s t h e favori te (alcoholic) beverage John drinks; as a n ex is ten t ia l habi tua l sentence, i t means t h a t John does not object t o drinking beer. Actually, t he re i s a third reading present , meaning t h a t John h a s t he habi t of drinking beer, without excluding t h e possibility t h a t he has t h e habit, of drinking wine as well.

Lawler assumes two different generic operators t o handle cases l ike th i s one; if there i s a third reading, we would need even th ree operators. But I th ink th i s assumption i s not necessary, a point which was made already by Spears (1974) for Lawless distinction. Firs t of a l l , not,e t h a t t h e three readings a r e associated with different intonation pat terns:

81.a) John drinks 'beer. (universal generic reading) b) John 'dr inks beer. (exis tent ial generic reading) c) John 'drinks 'beer. (habi tual reading)

From these intonation pa t te rns , we should expect t h a t in (a ) , only beer i s in t h e antecedent , i n (b) , only drinks, and in (c), both drinks and beer. Surprisingly enough, t h i s gives u s t he intended readings.

( a ) can be paraphrased as: On any pragmatically specified occasion where John drinks siimething, it I S beer by defaul t . (b ) s a y s t h a t if there i s a n occasion where there la some beer avai lable , John drinks i t by defaul t . Although th i s interpretat ion may hceiri too s t rong a t f i r s t s igh t , remember that t h e DEFAULT quant if ier weakens i t considerably, because there could be reasons why he doesn't drink bear, e.g. because h e decided to drink wine on an occ:ision where wine is a lso avai lable and there a r e rules s-lying t h a t John does not object 1 0 drinking wine and does not drink more than o u r beverage at. one occasion. Finally, (c ) i s a habi tual sentence of t h e type we have encountered earlier

1.s a Curther intere:->tuig example, kt u s consider another sentence of Lawler (1972):

82I.a) Delirier 'walks to school.

This sfinterice can be paraphrased, according Lo Lawler and avoiding t h e complications of t he DEFAULT operator, as: "on a l l occasions when Delmer goes to school, he walks". ( ( lear ly , walk nuist be considered a s p a n of t h e consequent, and th i s explains why i t get,s Lhe main sentence accent in t he reading of (82' .a) we a r e interested in . But in order L o r a r ry o u t this analysis , we have t o assume t h a t t h e meaning of walk can be dissociated into two components, t h e concept of moving and t h e concept of using one's feel, because the first component i s par t of t h e antecedent and only t h e second one is par t u f t h e consequent. Actually, it is t h e second component which a t t r a c t s t he sentence accent . This can be seen in languages which do not have a simple concept of walk, a s e.g. German; in these languages, it. is the second component which i s accentuated:

ri2I.b) Dehner geht zu 'Fup zur Schule.

Page 29: An Outline of Genericity

In t h e formalization of English, I assume a complex representat ion of walk, namely on-footl(go'). Then (82I.a) ge ts t h e following interpretat ion:

If t h i s analysis i s basically correct, i t means t h a t t h e assignment of semantic material to t h e antecedent and to t h e consequent i s extremeley free. I t not only cu ts accross syntac t ic borders, bu t even breaks open t h e semantic representat ion of words. This could be considered as problematic, because syntax and semantics seem t o be r a the r detached from each other. But note t h a t t he accent s t ruc ture res t r ic t s t he possible representat ions considerably. This simply means t h a t semantic interpretat ion has t o t a k e in to account no t only t h e syntac t ic s t ruc ture , but also t h e accentual s t ruc ture of an expression.

As another example, l e t u s consider re la t ive clauses in I-generic sentences. I t i s well known t h a t expressions containing r e ~ t ~ r i c t i v e and non-restrictive re la t ive clauses differ in Uieir accentual s t ruc ture , a fac t reflected in English punct,uation. Basically, a nori-restrictive relat ive clause i s a predication of its own on a well- established discourse referenl., whereas a restr ic t ive relat ive clause i s only par t of a predication and serves t o single ou t a discourse referent. Now, 1-generic sentences can have only restr ic t ive relat ive clauses except if the re la t ive clause i tself is to be considered a s generic, a f ac t discussed in Burton-Roberts (1976) and a t t r ibu ted to Smith ( 1 96 1 ):

riijl.a) A whale which i s ill 1 yields no blubber. b ) * A whale, which i s ill, yields no blubber. c ) A whale, which is a mammal, suckles i t s young. d ) " A whale which i s a mammal i suckles i t s young.

I t is bound to nbsume thal, restr ic t ive relat ive clauses a r e par t of t he antecedent . O I I H reason for t h i s is t h a t there is no optional pause between t h e indefinite NP and l l t i b rclat i i e claust1 Ivon-resLrictivo relat ive i-lauses, on the other hand, can be :irtalyst-11 H S par t s o f t,he consequent, a reason for th i s assumption i s t h a t they a re titipariilf~d From I t i t i i u i e f j n i k ~ NP by 'in opt iunal pause and contain a rhematic c o n s t i t u e n t B u t flien Hie distribution pa t te rn of (83') i s predicted.

Firs t , ( a ) is o l( because the relat ive clause can b e applied to t h e predicate in t h e : i i i icc~' i ic i~ i ( b ) I & bad because 1.he relaLive clause must b~ part, of t he consequent, i i i d t h u s a getiprn' s t a t e~~uen t about whales i s made which is obviously false , namely 1 h a t a wh:ile is i l l I j y defaul t (c) , on the other hand, makes a st,atement, about whales whii1h is o t .v ions ly < r u e . and therefore i s 0.k. ( d ) , finally, is out because t h e informalion o f I he relat ive clause does no1 restrict, t he antecedent in any way ~ i n i therefore fail5 t o rneel Hie basic pragmatic requirements for res t r ic t ive relat ive V ~ ~ I ~ S Y S .

b'itiall>, the re is this well-known example which once was designed t.o prove t,hat t h e passive transformation does not preserve meaning:

89".a) Beavers ! b u i l d 'dams. b) Dams i ace built, by 'beavers .

Clearly, these t w o sentences h:-ive a ditl't.-roil! accentual s t ructure. And i t is th i s difference which arcourits for their meaning difference i n meaning. We get t he following interpret at i i j r ib a.cr+irdiiig 1 o our l h o o r b .

Page 30: An Outline of Genericity

The f i r s t sen tence expresses a defaul t rule for beavers, and t h e second one, for dams. This is a n in tu i t ive correct account for t h e meaning difference of (83It.a,b). Clearly, then , these examples do not prove t h a t passive does not preserve meaning, because the i r themehheme-structuring has not been kept constant . I t indicates , however, t h a t passive of ten is employed, together with intonation, t o mark the themehheme-structuring of t he sen tence .

1 will now discuss short ly generic anaphora. Consider t h e following examples:

83.a) John killed a spider1 because theyi a r e ugly. b) John didn't keep a spideri because theyi a r e ugly.

The na tura l reading of t he second clause of (83.a,b) i s t h a t generally, spiders a r e ugly. They should be analyzed as referring to a kind; for example, i t allows for t he interpretat ions typical for kind-referring NPs, a s in John killed a spider because they once frightend his girlfriend. Generic anaphora of t h i s sor t can be explained if we assume t h a t a n NP with a common noun in any case introduces the kind the common noun is re lated to (an idea developed recently by Frey and Kamp).

In the examples we considered so fa r , t h e generic pronoun is plural. But i t can also be singular, as i n (84.a). If t h e antecedent i s plural, however, t h e pronoun must be i lu ra l as well, as m ( b l :

à § a ) John shot, a lioni, a l though i t s l i s protected. b) J o l i r ~ s ho t lionsi, although they1 a re / *it1 Is protected.

I't11.s r a n \ it \ e x p l a i n s 1 by three interact ing principles. The f i r s t one i s syntact ic : If t h e u n t ~ c e t i e n i is plural, i .e . ha s a marked agreement fea ture , then t h e pronoun must t ~ , a r the suine fea ture . The second one i s t h a t reference to a kind i s possible with p l u r a l N i ' s (cf t he use of bare plurals a s kind-referring terms). The third principle .says t h a i in cases where a singular pronoun could refer t o both t h e individual ent i ty : ind tile k i n d , a plural pronoun i s chosen t o refer t o t h e kind But if i t i s clear for I J ~ ! I ( - Â ¥ ceasons Uiiit the pronoun refers t o t he kind, e.g. because it, i s a n argument of a kind predicate a s in (84.a), then singular pronouns a r e also allowed. The third prtiiciplt', 1 hen, se rves to disa~nbiguate between different readings.

Miolher case of generic anaphora a r e indefinite pronouns like some and one (or German wlch --), as exemplified in

85.a) John saw a spideri , and Mary saw onei, too. 1 1 ) J o h n bought inilk,, and Mary bought somei, too.

8 6 a) OtLo sail erne Spirirtei, und Anna s a h auch einei. D) QUO kaufte Milchl, u n d Anna kauf te such welchei.

Inde f in i t e pronoud~ ran be analyzed a s referring to a kind which i s introduced in the preceding context , and a s introducing a realization of t h i s kind. Thus, they share propertit?& of definite and indefinite expressions.

rhert! are cases i n which indefinite generic N P s r an be antecedents of anaphora a s well:

8 7 ) A l i n o i i s ;i fftrocious beast. Iti has huge claws.

Page 31: An Outline of Genericity

I t i s c lear how t h i s example can be interpreted along t h e l ines t h a t we argued for above: Take a lion, and i t will be a ferocious beas t by default . Moreover, t h i s lion will have huge claws by defaul t a s well. This means t h a t t h e second sentence has to be interpreted aga ins t t h e background introduced by t h e f i r s t , Tha t is , in interpret ing t h e second sentence, one must be able to "look into" t h e f i r s t one. Although th i s procedure may be problematic from a technical point of view, i t seems t o be t h e r ight one to handle sentences l ike (87).

5.6 Generic and Episodic Interpretation. oc- Bare NPs - -- - ---

The analysis of genericity 1 have argued for above cont ras t s with the s tandard theory developed in Carlson (1977). Basically, in th is theory bare NPs always denote a kind, whereas in my theory, they can e i ther denote a kind (in the i r D-generic use) or be interpreted a s predicates (in I-generic uses or in episodic sentences) . In th i s section, I will discuss t h e distinction between bare NPs interpreted genericall) (be it, D - or I - ) and bare N P s interpreted in episodic sentences, which i s blurred by t h e s tandard theory.

A ra ther straightl'orward t e s t to s ee if t h e s tandard theory is on the r ight track i s provided by examples like t h e following ones, a s discussed by Schubert & Pelletier ( 1 987):

88.a) Snow i s whit,o and is falling r ight now throughout Alberta. b ) Dogs a r e mammals and are barking right now i n f ront of my window

Schubert. & Pelleticr think that examples l ike these a r e 0.k. My impression i s t h a t t h e y m e actuaily ra ther bad. That they a re accepted by some speakers can be o ~ ~ i l i i i n e d i t " hts au:-iitifie that ihese speakers apply tohe kind-oriented mode of speaking (r l ' st:vtioii ,i i-i). (11 l lns case, botti co t~ juncts of t he examples in (88) would be [ ~ r e ( l i r ; ~ l i.*(l to a kitid.

' l i f - I I I ~ L I J I retison w h y (2irlaori developed his theory in the f i rs t place was t o explain ~ l . y tit^ interpretat iun of bare NPs a s generic or non-generic depends heavi ly on the verbal [.imjit+?it~?. According to him, hare plurals always refer t o kinds, and episodic verbal predicates reduce a proposition about a kind to a proposition about a t ' t - ~ : i l i ~ , ; t ~ i o n u f a kind C:irlson's analysis capt . im~s the assumptions that

1 ) ( l i e ir i terpretaiion of bare NPs depends on the verbal predicate, I i j hare I L K S typically have narrow scopi- when interpreted nori-

possi l ) l ~ bd ween generic and nor1 -

Carlson 's claims with r eaped to ( i i )

ll'or c l a i i o ( i i i ) , c ; t t l so r~ ' s theory falsely predicts t h a t t he following sentence can get a decen t interpretat ion.

Accorctiiig t,o Carlson's theory, t he hniding of t h e m by app les should be perfectly pussible i n a normal readirig of t,his sentence, because both NPs refer ti) t h e same en t i t y , a kind, and the verbal predicate a te serves lo single out realizations of \\ns kind whicli a r t 1 posh ib l~ different from t'sich other . B u t note t h a t (89) actual ly sounds ra ther str<iiipth There could be a reasonable interpretat ion in t,he kind- oriented mode o l s ~ ~ c ; i k i t s p , , i . ( : i f t~otli sent.ences a re interpreted a s involving the kind apples a r n ~ iiavt: the representat ive object reading. 1 think i t i s again th i s

Page 32: An Outline of Genericity

reading which i s avai lable for speakers who do not find (89) so bad.

For claim (ii) , Carlson's theory fai ls t o give the right coreference conditions in cases l ike t h e following:

9 0 ) John saw applesi on t h e plate , and Mary saw themi, too.

In t h e most na tu ra l reading, apples and them refer t o t he same objects. According t o Carlson, t hey refer to t he same kind. and i t i s only t h e predicate saw which introduces objects. But a s t he ex is ten t ia l quant if ier has necessarily narrow scope (because i t i s introduced internally in t h e lexical semantics of t he verb) , there i s no means t o express t h a t apples and them should involve the same apples. (A similar point was made by Kratzer 1980).

Furthermore, t he re a r e languages which do distinguish between generic and non-generic bare NPs. For example, in Finnish a generic bare NP i s in t h e nominative case, whereas a non-generic bare NP is in the par t i t ive case:

9O9.a) Maito on makeaa. 'Milk (MOM) i s sweet . b) Maitoa kaa tu i p o y d a l l e . 'Milk (PART) was s p i l l e d over the table'

1 have suggested a n a l te rna t ive theory which assumes t h a t generic bare NPs a re analyzed e i ther a s indefinites in t h e scope of a defaul t quant if ier , or a s proper names of kinds, and t h a t non-generic bare NPs a r e indefinites which a re not in the scope of a defaul t quant if ier .

-1s hare NPs a re analyzed similarly t o other Indefinites, they should exhibi t the same scope behavior a s other indefinite NPs. This explains why wide scope bare NPs a re possiblt-'. B u t cer tainly, bare NPs a r e more of ten interpreted with narrow scope than indef in i t e N P s with a determiner. One reason probably i s t h a t a specific, i .e. wide- s ~ o p e , roathri}; of ~ridtifimt~eti is the inort; favored, the richer t h e descriptive content uf the N P is. A s bare NPs have a poorer descriptive content than other indefinite Nils, s imply because the number or quant i ty of t h e en t i t i es they refer to i s l e f t i i n s p ~ ~ i t ' ~ u d th is explains why they lend to be interpreted ah non-specific, i.e. with narrow ss.'opn

l i u l w h ) dues the na ture of the verb have Lhis s t rong impact on t h e interpretat ion of a serii,unc^ in-, be ing generic or non-generic? 1 think we need Lo explain two things:

1 1 W l i k ( t o 1--gt/ntiric sentences always have s t a t i v e predicates? i i ) Why d o sLali\ e vnrhs not allow for bare NPs a s subjects in non-generic sentences?

I 1 c u l i i ~ - ~ n ( 1 ) we liii\/e already solved above u-C. section 2.2). We showed t h a t , a s I- generic s f~ntcnces quantify over occasions, they cannot be episodic, bu t have to be ~ I i i t l V ~

'rotilmu ( i i j shou ld he associat,ed with t h e research on accented subjects (e.g. Fuchs 1980, Clussentioveii 1983). I t i s well-known that certain verbs allow for accented, non- I

' o n 1 rt i i i l ivr s u b . l ~ c l s, whereas others do nut , 1 9 1 . a ) 'Jutiti has arr.1 v e d from Ireland.

h j A ' m a n has arrived from Ireland.

9-i.a) * ' John l ives i n Ireland. (only possible when b) ' A 'man l ive? . in lrelund con l rast.ive)

Note that. the fol lowing s p h i r ~n ' e s , which differ in intonation, a r e 0.k. I

Page 33: An Outline of Genericity

Sentences l ike (91) a r e called thetic, and sentences l ike (93) a r e called categorical, a terminologiy developed by Brentano and Marty in t h e l a t e 19 th century and revived by Kuroda (1973); s e e also Ulrich (1982) and Sasse (1987) for a survey of t h i s distinction in a number of languages.

Although i t i s s t i l l unclear what exact ly determines whether a verb can occur in thet ic sentences or not (I w i l l not go into t h e different solutions proposed in t he l i terature) , i t seems a t l ea s t clear tha t stat,ive verbs a r e hardly ever found in thet ic sentences. Of course, th i s holds only if we exclude sen tences expressing t h e current location of a n object, l ike John is in the room, which a re sometimes called s t a t i v e a s well.

We can n o w recas t problem ( i i ) and ask why bare NPa hardly occur a s subjects of categorical non-generic sentences, although they a re acceptable a s subjects of thet ic sentences, a s shown in the following examples:

94.a) 'Paintings arr ived from Ireland. b ) 'Whisky arr ived from Ireland.

95.a) ?? Paintings ! a re in Ireland. ( in non-generic b) I!? Whisky ! is in Ireland. i r~ t~erpre ta t ion)

A n explanation of t h i s fac t can be given if we consider t he different functions of the verbal predicate in the t ic sentences and categorical sentences. The main function of t he verbal predicate in thet ic sentences i s t o introduce a new en t i t y in to t i i~ (~o i i r s e This assumption can be corroborated by t h e f ac t t h a t t hese verbs normally (jest-nbe manners by which a n entit,y comes into being, dissappears , and generally a t t r a c t s the consciousness of par t ic ipants . These verbs have, loosely speaking, a built-in e x i ~ t ~ e n t i a l quantifier for the i r subject referents , and I call them therefore exis tent ial verbs O n Die other hand, t he verbal predicate in categorical ser i f t-ficcs s i m p l y ib pt'e(iit:%t oci to i tic bubjoct re feren t and has no ex is ten t ia l quant i- in-'ta b m l t intu i t Theratfore, i t must apply ti) en t i t i es which a r e already given a t

t i t ) I iin ojil discoiirt7c siaf i?, t ha1 I S , L o I-li t if i i e s whicti have been introduced earlier in U K disi onine or which a r e part, of' t h r shared knowledge of speaker and hearer . In i in>hi (Â¥;ii-nl.- 1 fit:; aii-; in fact nit.ro~iticcii varlier in discourse. B u t i t i s possible tha t I 1 . 1 1 a111l t i n : hu~t ic s e n t e n c ~ , an NP introduces a n en t i t y and a categorical verbal p t t J t l i i I ; ( is applied t o i t ( S i ; i I ) ) 1:; :HI t: la~riple a t hand

N O W 1 1 i.-cins t h a i i i ' an en t i t y has to t i t A introduced into t h e discourse by a n NP : i l ~ i r i i - w t i t t o~ i t the he lp o f a n exis1,ential verb, t he NP should be considerably complex 1 3 O ( , i n 1 5 ,101) prop:-t i > it should contain a determiner, t h a t is, 1 1 should fml i ) ~ ;i h i i r ~ NP tint clearly, this i s only a preference rule, a s bare NPs can in pt iiiciplc nil t ~ i i i n ~ ~ ~l i scourse referents hy the i r own, for example in t h e language of I ' l f !WS[~K(i t ! f .S '

9 6 ) l '~ ut't'n*;al<j yesterday denied a n y involvement in the affair . They maintained 1 liat ( . . . I

1'hn.s A V have a n esplanaliori why a bare NP and a s t a t i ve vvrbal predicale cannot roiiitnin; (excepi , of course, w i th I-generic sentences) : The s t a t i v e verbal predicate doeb not have Hit- ( . . .~s t ,en i ia i force t o i i~troiluce a neai en t i t iy inio discourse, and i.tit? bare Nl-' does h a \ e i t only i n a inarqinaiiy.

Wliy ih i t f l i s i l bas") M's and i i ~ l e f h i t o hit's * i t h deterrriiners d i f f e r in t h i s respect? ' w o puhsihle re;isoiir> corne into mii i i i K n b s t I ! determiner of the indefinite NP could A - ~ s b u c i a l ~ d wi th some "iniri~duciiig" lorcc, l i k e an exis tent ial quant if ier . Bare NPs would l a r k hi:, mir-tjducmg force b ~ ~ a u s ~ t h e y lack a determiner, and would t,herefore rwvd the introdiicinp, f'orce of a t . h ~ l i c predicate. Secondlv, i t , could be t h e case tha t an indefinite i\P whirh intriiduces a new f i i t i ty by its own has t o be semantically spt.'i+ifh- enough, a n d we have seen l h a t bare MCr, lack specificity when compared with indefinite NPs with determirit;r.

Page 34: An Outline of Genericity

References

Bacon. R. 1973. Do Generic Descriptions Denote? Mind 82. Biggs, C. 1978. Generic Generalizations. Akten des XII. Kongresses fu r Linguistik,

Wien. Burton-Roberts. N. 1977. Generic Sentences and Analyticity. Studies in Language 1. Carlson, G . 1977. Reference t o Kinds in English. Ph.D. disser tat ion, Amherst.

Published 1980, New York: Garland Publishers. Carlson, G . 1979. Generics and Atemporal When. Linguistics and Philosophy 3 , Carlson, G. 1986. The Semantic Composition of English Generic Sentences.

To appear in G . Chierchia @.a. (eds.) , Property Theory, Type Theory, and Semantics. Reidel.

Chafe, W . 1970. Meaning arid t h e Structure of Language. Chicago: Chicago University Press.

Chierchia, ( 3 . 1982. Nomirializations and Montague Grammar. Linguistics and Philosophy 5.

Chomsky, N . 1975. Quest,ions of Form and Interpretat ion. In R . Austerli tz ( e d . ) , The Scope of American Linguistics. Lisse.

GroSt, H i l l . 1986. Universal quant if iers and generic expressions. Manuscript, Stanford University.

Dahl, 0 1975. On Generics. In E. Keenan (ed.), Formal Semantics of Natural Language. Cambridge: Cambridge University Press,

Dahl, 0. 1985. Remarques s u r Ie generique. Larigages 79. Ebert, K. 1971. Referenz, Sprechsituation und die bestimmten Artikel

in einem nordfriesischen Dialekt (Fering). Nordfriisk Ins t i tu t , Bredstedt. Karkas, 0 & Y Sugioka. 1983. Restrictive If/When clauses.

Linguistics and Philosophy 6 . F n r h s , A . 1980. Accented subjects in "all-new" sentences. Wege zur

t~iiiveraalic:r~forschurtg (Festschrif t f i ir Hansjakob Seller). Tubingen: Narr. i i i ~ l n u c I n i , M l 98f- H i ~ " i i b e ~ , syntagmes e t art ic les g6neriques Langage 79. Gcrsl ncr , (' i 9'/ 9 H r l w ( i eoe r i z i t a ( ?~ . M . A thesis , liniversit,y ol' Munich. 1 ~ ( ; u r t h i i 1 985 i~t-'iiorlcs. Journal u f Seniantir s 4 . ' i i ip t i i , A 1980. The Logic of common nouns. New Haven: Yale University Press. ~ . i i ~ s s i l ~ ~ i i n v ~ r i , i'. 1983, Focus, mode and the nucleus. Journal of Linguistics 19. in-nil, l 1982 The Serri~inties of Definite and Indefinite Noun Phrases.

1 I 1 d i s s e r t a t i o n , Amherst. Published by Sonderforschungsbereich der i i ~ i v t rs11,;it;t hurislanz, Nr. 73

H(>111, I 1987. A Note on Negative Polarity and Downward Entailingness. NELS 14. In w , (; 1985. (Jurit;ric DesrripUons, Default. Reasoning, and

T 2 pir:iI I t ) Tln'urel ical Linguistics. t t eyc i , Ci 1987 G P ~ I P ~ i;,(:ht; ht?r~rizeichi~inigeii 'Zur Logik und Or11 ologie

g ( ~ r i e r ~ s r h e r Bedf-'uturigen. Munehon, Wien: Philosophia. i ' i i -ht ncloi i , 12. I !!?:I Semantic Interpret a1 ioii i n a Generative Grammar.

( .dniM idge'ivlass i \ ; i i ~ i p , ( 1 1981 A Tlieury ol Truth arid Semantic Representation. In J.

(it'oerieiidijk e a (cds ), Formal Methods in t he Study of 1,anguage. Mullit-nuit t.icli t:t:iir,rinrl, Amsterdam

ha) , 1' 1 Y', I 'I axonoiny atui Semantic Contrast. Language 47. l - ' l c ~ t , ~ 1 , (; 1 985 O n I b6 tc O P la g ~ n e r i c i l v~ rba l t : : los approches

quanl ifictit iom I1e.s. L;iiigages 19'79 K r a l ~ ~ r A 1980. \)I(, Aiinlyse des blossen Jiliira1 hei Gregory Carlson

L,inguislische Bt~richte 70. KnfKa, M I !)ti7 Noniinal Koference und 'l'oiuporal Constitution, To

appear in I he 1'1 oceedings of t h e 5 th Amsterdam Colloquium. ~ ~ r o d a . , S . Y. 197:: Fhc categorical and t he I hetic judgement. Evidence

from Japanese syntax PoundaLions of Language 9. Ladusaw, B 197q. Ncg,cttivi- i'oliifity a s lnhereiit Srope Relations. Ph.D.

dissert auuii. Un ive r s i t~ of Texas ~ar igford , C H 1 'È4d 'I'he Institutional Use: of "The". t'hilosophy and

i't~enomenoiogical Research 1 0 . I,nwItir, J. 197: Gem vies l o ;i iaiill Ghicago Linguistic Society 8.

Page 35: An Outline of Genericity

Lawler, J . 1973. Studies in English Generics. Ann Arbor: University of Michigan Papers i n Linguistics, 1.

Lewis, D. 1975. Adverbs of Quantification. In E.L. Keenan (ed), Formal semantics of na tu ra l language. Cambridge: Cambridge University Press.

Link, G. 1983. The Logical Analysis of Plurals and Mass Terms: A Lattice- theoretical Approach. In R. Baeuerle e.a. (eds.) , Meaning, Use and the Interpretat ion of Language. Berlin.

Newton, B. 1979. Scenarios, modality and verbal aspect i n Modern Greek. Language 55.

Nunberg, G. & C. Pan. 1975. Inferring Quantification in Generic Sentences. CLS Papers 16.

Partee, B. 1973. Some Structural Analogies between Tense and Pronoun in English. The Journal of Philosophy 70.

Pelletier, F.J. & L.K. Schubert. 1987. Problems in t he Representation of t h e Logical Form of Generics, Bare Plurals, and Mass Terms. In E. LePore (ed.), New Directions in Semantics. London: Academic Press.

Perlinutter, D.M. 1970. On t h e ar t ic le in English. In M. Bierwisch & K. Heidolph (eds.) , Progress in Linguistics. Den Haag: Mouton.

Plat teau, F. 1980. Definite and Indefinite Generics. In J. van der Auwera (ed.), The Semantics of Determiners. London: Croom Helm.

Reiter, R . 1980. A Logic for Default Reasoning. Artificial Intelligence 13. Sasse, H . - J , 1987. The thetic/categorical distinction revis i ted. Linguistics 25. Smith, C.S. 1961. Determiners. MIT Research Laboratories in Electronics,

Quarter ly Progress Reports 63. Spears, A.K. 1974. On t h e notion occasion and t h e analysis of aspect .

Chicago L,inguistic Society 10. Strighi, A 1986. Eine Semantik fuer generische Saetze (am Beispiel des

Englischen). Doctoral Dissertation, Akademie der Wissenschaften der DDR. t,er Meulen, A . 1986. Generic information, conditional contex ts and

eonslraints . In K . Traugott e.a (eds 1, On Conditionals. Cambridge. Turner, R . 1983. Mnnlague Grammar, Nominalizations and Scott 's

i)oinairis. Linguistics and Philosophy 6. I ~ i r i c h , M. i 98'2 Thetisch und kategoriw h Tubingen: Narr. \ ftltnit'm, F" 1987. Data Semantics and t,he Pragmatics of indicat ive

roi1111t ~or~itLs 111 F. Traiigott e .a . , O n Conditionals, Cambridge: ( ; : , i i ibr i t i g~ O i i i ~ ~ r s i t y Press.

\i ~ i r ~ t l l i ~ ~ ~ , Z. 1967 1,inguistics and Philosophy. Ithaca: Cornell University Press. , , :.(;niach, F . 197U. Voijr Ontologies. Journal of Philosophy 67.


Recommended