A unified theory of ((in)definite) descriptions

Transcription

1 Gennaro Chierchia University of Milan - Bicocca Amsterdam Colloquium 2001 A unified theory of ((in)definite) descriptions (1) Quantificational (denotation of type <<e,t>,t>; e.g. every man) NP?? indefinites Referential (denotation of type e; e.g. the man) I. Data. (2) What is an indefinite? If we take the indefinite article as representative, then indefinites appear to be characterized by the following properties: a. They have an existential meaning ( there is a boy such that.../ for some boy,... ) b. They admit of predicative uses. i. copular constructions: John is a nice man John and Bill are two nice guys ii. the Det itself admits of a quasi adjectival use They are two John is only one [The kind of things we are sure to be indefinite articles are nearly always derived from the first numeral] c. They support existential constructions. There is a boy in the garden d. They have generic (quasi universal) uses. a dog barks e. They display Quantificational variability effects. Donkey anaphora i. If a cat has blue eyes, it rarely is intelligent ii. A Texan is often tall f. They support discourse anaphora A boy walked in. He... (3) Some-indefinites. Some indefinites (e.g. some) have a (very small) subset of the properties in (2). (a) No (or marginal) predicative uses. i.?? John is some nice boy ii. * John is some (b) No (or marginal) in existential constructions. *There is some boy in the garden (c) No (or marginal) generic uses. some dog barks (d) No (or marginal) QV effects.?? if some cat has blue eyes, it normally is intelligent?? some Texan is rarely short Some-indefinites seem to have only (2a) and (2f) in common with the indefinite article. (4) The scope of indefinites. Indefinites, in general, allow scope construals much more permissively than other quantified NPs a. Every student read every paper that a/some professor wrote b. every student i some professor j [t i read every paper [ that t j wrote ]] (5) Certain-indefinites Modifiers like certain bring to salience wide scope or specific readings a. Every student read every paper that some professor wrote Not every specific construal can be reduced (in an obvious manner) to scope b. (At the bachelor party) no one invited a certain girl i. Namely Melissa (specific individual reading) ii. Namely, his fiancé (specific functional reading) c. No one invited a/some girl [only reading (b.i) is present] d. No one invited somebody he should have invited

2 II. Definites. (6) Strategy To find our way between quantificational and referential uses, let definites be our guide. What makes definites referential or specific is not simply that they refer to individuals. a. Everyone met the leader b. Everyone met the leader of his group (7) The classical theory a. the boy = i(boy) = the unique member of BOY b. the boys = i(boys) = the maximal member of BOYS the is a choice function with a maximality presupposition; where does such a presupposition come from? (8) A slight (?) variant of the classical theory (Chierchia 1995) a. the denotes a variable over choice functions b. the f boy arrived ==> arrive (f(boy)) If this is what the means, for communication to be successful the context will have to supply a unique (or a maximally salient) value for the function variable (much like it has to supply a unique value for he in he smokes). If there are two or more (equally salient) boys, then there will be two or more (equally salient) choice functions; and communication failure will ensue. It follows that we can use the boy only in contexts in which there is one (or one maximally salient) boy. [c. Othere definites: THAT(BOY) ==> select from BOY the one pointed at ] (9) Domain selection and/or hidden parameters a. The kid is tired [doesn t presuppose a unique kid in whole world] b. TIRED(i D (KID) ) DOMAIN SELECTION c. i. The f kid here HIDDEN PARAMETERS f(kid)(here) d. DP D NP n D NP The f boy there n f (KID) (THERE) (10) Anaphoric associative uses: an advantage of the hidden parameter strategy? a. Every student read a book and reported on the author c. Every student read a book i and reported on the D(i) author d. Every student read a book i and reported on the f author i (11) Consequence: Weak Crossover effects By assimilating the hidden parameters to covert pronouns, we expect them to obey to the standard conditions on pronoun interpretation. a. * His i mother loves everyone i b. Everyone s i mother loves him i c. Everyone that read a book i selected by our panel was very impressed by the f author i d. * The f author i impressed [everyone that read a book i selected by our panel] Other relevant effects: VP ellipsis, association with focus. (12) Covert descriptions: laziness pronouns a. John wrote only one paper. It was on adverbs (Heim&Kratzer) i. i (PAPER THAT JOHN WROTE) [ traditional ] ii. f( paper)(john) [Choice function] b. Laziness pronouns need a syntactic antecedent. Salience is not enough. Coindexing = maximal non distinctness (copying/interpretive procedure) Typically, the head of the antecedent is interpreted as the argument of the choice function.

3 John i wrote only [one paper] j. It j i was on adverbs it = f antecedent = [one paper] j. It j = f(paper) c. Every student i who wrote only [one paper] j presented it j i twice d. DP DP DP D NP pro i It f Ø j f P x i e. Paycheck pronouns John gave his paycheck i to Sue.[ Bill/everyone else] j gave it i j to Mary (13) Crossover effects on covert descriptions. a. * It i j was presented by every student j who wrote only one paper i b. Everyone got something for the holiday season. Only Bill was pleased by his present. c. i. Everybody else hated it ii.?? It irritated everyone else (14) The proposal. a. Definite descriptions are choice functions with hidden parameters (Skolemized CFs) b. Indefinites denote the same thing as definites c. Indefinites must be closed existentially ( morphological requirement ) (15) How to interpret it a. One new alternative source of indefiniteness: every indefinite is a choice function (subject to closure) [globally responsible for all their properties] b. Two sources of indefiniteness: i. GQ over individuals (e.g. lp lq $x [P(x) Ÿ Q(x)]) [responsible for quantificational uses] ii. E-closure of SCF [responsible for referential construals] Given the widely diverging properties of indefinites (e.g., a vs. some), I am inclined to side with those who claim that the sources of indefinites are (at least) two. (16) Existential operators. a. NOT directly associated with a lexical item. b. Inserted (freely) in a structure that contains indefinite morphology. c. They come in two forms: i. $ = lp lq $x [P(x) Ÿ Q(x)] (over individuals, via sortal properties) ii. $f= lq $f [SCF(f) Ÿ Q(f)] III. Implementation (17) a. Type driven translation (over SCFs) DP John e VP smokes <e,t>

4 b. Indefinites DP VP smokes D man <e,t> <e,t> a if a is just a indefiniteness marker; man and smoke cannot directly combine. This yields two options. (18) Option 1 a is interpreted as a choice function, then the type will match. $ f SMOKE(f(MAN)) [I am omitting the implicit Skolem argument for simplicity] DP VP smokes a man f However, existential closure will have to apply after the clause nucleus has already been formed. I.e. non locally. (19) Option 2. We insert the existential quantifier locally DP VP smokes D man <e,t> <e,t> $ a $ (MAN)(SMOKE) (20) Some-indefinites are lexically specified to be functions: some f a. In copular constructions ( John is some boy ) they cannot receive a predicative interpretation b. *There is some boy ==> Exist (f(boy)) ; if defined is tautologous c. No (or marginal) generic uses i. a dog barks ==> Gn x, [dog(x)][bark(x)] ii. some dog barks ==> * Gn f [bark(f(dog)] d. QV effects are linked to topicality. Functions make poor topics i. * As for one s mother, she is a source of trouble ii. As for one s mother, one should never let her take over (21) Certain-indefinites are a way to disambiguate a in favour of the functional construal. a. DP i f(man)(i) D man a certain f b. Ogni studente ha letto ogni libro che un qualche professore ha scritto. Every student read every book that a some professor wrote c. A certain also has a prominent kind reading i. A certain dog is hard to train ii.?? some dog is hard to train iii. A certain argument is rarely convincing iv.?? some argument is rarely convincing d. Gn x i [x i is an instance of k(dog)(l)] [x i is hard to train]

5 (22) Choice functions and presuppositions a. A = [U Æ 2] (the set of all partial functions from U into {0,1}) b. We quantify over SCF that are not defined for the empty set: SCF = {f Œ [AÆ[U Æ U]] : œdom(f) and for any aœa and uœu, f(a)(u)œ a } c. The closure operator is a generalized quantifier with a built in restriction to SCFs: $f D = lq $f [D(f) Ÿ Q(f)] (D some natural subset of salient SCFs) d. Example 1: some man walks i. $f [some f manl walks] (l = the context location) ii. $f [SCF (f) Ÿ walk (f(man w0 )(l))] ( $x[man(x) Ÿ walk (x)]) [If there are no men in l, (ii) is false] e. Example 2: some man doesn t walk [two scope construals -- putting aside alleged Pos Pol] i. $f [some f manl ] i not [ t i walks ] Ë $f [SCF(f) Ÿ ÿ walk (f(man w0 )(l))] ( $x[man(x) Ÿ ÿ walk (x)]) [If there are no men (or if every man walks), this is false)] ii. not $f [some f man l walks ] Ë ÿ $f [SCF(f) Ÿ walk (f(man w0 )(l))] ( ÿ $x[man(x) Ÿ walk (x)]) [If there are no men (or if no man walks) this is true] Normal behavior of indefinites (more context dependency than on the classical view; but very weak presuppositional requirements). IV. Consequences. (23) The scopal properties of indefinites a. $ f [every student i read every book that some f professori wrote] b. every student i $ f [t i read every book that some f professori wrote] (24) Crossover effects on LD construals a. a technician inspected every plane b. every student was impressed by every book that some professor wrote c. every book that some professor wrote impressed every student No intermediate reading d. Every book that [some f professor] i wrote impressed every student i e. every student i [every book that [some f professor]i wrote impressed t i ] (25) De dicto/de re. The relevant contrasts can be derived on the basis of independently motivated assumptions. Assumptions: Each expression has a world variable (as in Ty2 translations a la G&S). The world variable of the main predicate of a clause must be locally bound (Percus); other predicate variables can be freely set (modulo pragmatic plausibility) a. John believes lw[it rains w ] Ë a. BELIEVE w0 (j, lw RAIN w ) b. De dicto/transparent reading [context: John finds his car scratched and sees a rowdy group of youngsters nearby: He believes one of them did it. They happen to be the local soccer team] c. John i believes w0 lw $f [a f footballplayer w0i scratchedw his car ] c. BELIEVE w0 (j, lw $f [SCR w (f(footballplayer w0 )(j)) ]) d. De re: $f John i believes w0 lw [a f footballplayer w0i scratchedw his car ] d. $ f BELIEVE w0 (j, lw [SCR w (f(footballplayer w0 )(j)) ]) There is some function f mapping John into a football player such that John believes that he scratched his car

6 (26) Donkey anaphora effects. Assumptions: donkey anaphora derives from characteristics of $ (plus independently needed assumptions) i. $ brings to salience an entity; pronouns are covert descriptions that can refer to salient entities [plus situation theory] ii. $ is dynamic Consequence: LD indefinite should behave like ordinary ones with respect to Donkey anaphora. a. John believes that a student in his class is a spy. She is from Rome b. John read every book that a professor has written. He wants to ask her to be his advisor c. every student that read every book that a professor wrote asked her to be his supervisor d. Most linguists that studied every solution that some problem had became crazy over it e. If John forms the opinion that a student is intelligent, he or she is indeed typically quite smart f. If John thinks he may need a credit card, he will carry it along g. If a student reads every paper that a professor wrote, it s because he wants to have her as advisor (28) A bug? a. every book that some professor wrote impressed every student b. every student i [every book that [some f professor]i wrote impressed t i ] c. every student i $f [every book that [some f professor]k wrote impressed t i ] (29) The immediate scope constraint (ISC) Between the binder of k and the existential quantifier, there is an intervener Closure of a SCF f must take place in the immediate scope of the binder of f s argument (i) * Q k Q i $f f(k) (ii) * $f Q k...f(k) (30) Some consequences of the ISC a. every student read every book that some professor wrote b. *$f every student k read every book that [some f professor]k wrote c. every student k $f read every book that [some f professor]k wrote (31) Schwarz s (2001) result. With wide scope closure (a) is predicted to have a reading equivalent to (b) a. Not every student read a (certain) book I recommended (namely, his assignment) b. Not every student read every book I recommended c. $ f [ not every student i read [a f book I recommended] i ] d. i. $ f [ [ d a ] l i [ [f b] i g] ] ii. [ [ d a ] [ [every b] g] ] whenever d is DE So, the ISC prevents overgeneration in two contexts: i. WCO contexts ii. DE contexts (32) Summary a. Proposal: referential indefinites have the same meaning as definites (SCFs), modulo closure (subject to ISC) b. Consequences i. LD construals of indefinites ii. Crossover effects on LD indefinites iii. De re construals of indefinites iv. Donkey anaphora effects on LD indefinites (33) Alternatives. a. Relaxing for indefinites constraints on movement. (Geurts) b. Domain selection: sometimes the domain of an indefinite is just a singleton (Schwarzchild) c. Indefinites as free SCF. No closure. (Kratzer) Problems with (b)-(c) d. Not every linguist studied every solution that some problem has. Charlie didn t. e. every linguist i studied every solution that some D(i) problem has f. not every linguist i studied every solution that some D(i) problem has g. There is no way of choosing a D such that Charlie studied every solution to some problem in D

7 (34) A further consequence: Belief de notione a.... a f witch... she f(witch) coindexing = non distinctness; if the antecedent is a function, this entails that a pronoun anaphoric to it will pick up the same function. b. Existential closure and world selection are independent processes. Hence the world variable of an embedded indefinite may be locally bound (e.g. within the scope of an attitude verb); while existential closure may apply at the top. This predicts the existence of a much discussed anaphoric dependency c. $ f [Hob believes that lw[a f witch w blighted his mare]. Nob believes that lw[she f(witchw) stole his pig]] d. $ f [ BELIEVE w0 (Hob, lw [BL w (f(witch w )(x)) ]) Ÿ BELIEVE w0 (N, lw [SP w (f(witch w )(x)) ]) e. There is a witch-concept such that Hob believes that whatever satisfies it blighted his mare, while Nob believes that whatever satisfies it stole his pig. (35) What is our witch-function a function of? In general for attitude reports we want to go for functions from the believe holders into the relevant concepts. So in (a) we might set x to one of the relevant belief holders. b. $f [ BELIEVE w0 (Hob, lw [BL w (f(witch w )(H)) ]) Ÿ BELIEVE w0 (N, lw [KL w (f(witch w )(H))]) (Straightforward pragmatic considerations would make us pick the first mentioned belief holder: you wouldn t make the function depend on somebody which has not entered the scene) This reading makes Nob s beliefs depend on Hob s ones (asymmetric reading) c. Typical concepts that would fit the bill: the witch Hob believes to have blighted his mare the witch that Hob believes to be around d.it is also possible, as Geach himself pointed out, to use the relevant sentence in a situation in which Hob s belief do not depend on Nob s. This is the so called witch-mania context. In this case, x in (b) could be interpreted as a collective agent (say, the community of which Hob and Nob are part). A suitable description could be: the witch Hob s and Nob s community believes to be around In this case, we have a symmetric reading. (36) A sloppy reading? a. Hob i believes that [a f witch] blighted his mare. Nob j believes that she witch j killed his cow b. Hob and Nob each have their own witch-concept; this reading seems to be impossible (imagine Hob and Nob living in different communities with no causal links to each other). (37) The case of Fred and Charlie (Edelberg (1986)) a. Fred s wife, Janet plays a practical joke on Fred. She makes him believe she bought him a new car, which is awaiting for him in George s garage. Charlie s wife, Mabel, plays the same joke on Charlie. She makes him believe she bought him a new car, which is awaiting for him in George s garage. Belief holder Belief Fred Janet gave him a car which is in George s garage Charlie Mabel gave him a car which is in George s garage b. Fred believes a new car is in George s garage and Charlie believes it is a gift from Mabel c. Fred believes a new car is in George s garage. Charlie believes that the new car in George s garage is a gift from Mabel. d. Charlie believes that his new car is George s garage. Fred believes it is a gift from Mabel. (38) Generalization on sloppy construals of pronouns of laziness. Sloppy construals of pronouns with definite antecedents is possible; with indefinite antecedents is impossible (or systematically harder). (39) The ISC again (Misani 2000) a. $f [Fred i believes a f new cari is in George s garage. Charlie j believes it f(car)j is gift from Mabel] The only solution is to leave the functional pronoun free. But a functional pronoun needs a linguistic antecedent. Contextual salience is not enough: b. Every donkey owner beats it c. Speaking of the successor function, no number is greater than it (Heim) d. Charlie believes that his new car i is in George s garage. Fred believes he got itnew car j as a gift from Mabel

8 (40) More analytically. Suppose we try to do things respecting ISC: a. [Fred i $f believes a f new cari is in George s garage]. Charlie j believes it f(new car)j is gift from Mabel] Fred i = $x i [x i = Fred It would seem that f(car) in the second clause is free. But it is not so. Given its dynamic character, $f winds up binding f(car) in the second clause. Hence a violation of ISC comes about in any case. b. Prediction: replace Fred i with a quantifier like every man i. This quantifier creates a barrier for $f (every is externally static; cf. *Every one has a donkey. It gets beaten regularly). Hence, f(car) in the second clause will remain free (while, at the same time, having a linguistic antecedent, that fixes the argument of f). Thus the possibility of getting a sloppy reading should increase. c. Everyone in this archaic community believes in guardian angels that assist people in the key junctures of their lives. In particular, every man believes that a guardian angel protects him at war and every woman believes that it protects her during child birth. d. i. [every man i $f believes a f guardian angeli protects him while at work] and ii. every woman j believes it f(guardian angel)j protects her during child birth OUTCOME: a deflationary theory of intentional identity (41) Summary [cf. (32)] a. Proposal: referential indefinites have the same meaning as definites (SCFs), modulo closure (subject to ISC) b. Consequences i. LD construals of indefinites ii. Crossover effects on LD indefinites iii. De re construals of indefinites iv. Donkey anaphora effects on LD indefinites v. Intentional identity Selected references Chierchia, G. (1995) Dynamics of Meaning, U. of Chicago Press Chierchia, G. (2001) A Puzzle about Indefinites, in C. Cecchetto et al (eds) Semantic Interfaces, CSLI Edelberg, W. (1986) A New Puzzle about Intentional Identity, Philosophical Review, 15. Geach, P. (1967) Intentional Identity, Journal of Philosophy, 5. Geurts, B. (2000) Indefinites and Choice Functions, Linguistic Inquiry, 31. Heim, I. (1990) E-Type Pronouns and Donkey Anaphora, Linguistics and Philosophy, 13. Kratzer, A. (1998) Scope or Pseudoscope? Are there Wide Scope Indefinites in S. Rothstein (ed.) Events in Grammar, Kluwer. Misani, N. (2000) Identità Intenzionale e Grammatica dei Pronomi, Tesi di Laurea, University of Milan Reinhart, T. (1997) Quantifier Scope. How Labor is Divided between QR and Choice Functions, Linguistics and Philosophy 20. Schwarz, B. (2001) Two Kinds of Long Distance Indefinites, this conference. Schwarzchild, R. (2001) Singleton Indefinites, ms. Rutgers U. Winter, Y. (1997) Choice Functions and the Scopal Semantics of Indefinites. Linguistics and Philosophy, 20.