Zur Semantik der Spezifizität bei indefiniten DPs und Satztopiks Cornelia Endriss Institut für Kognitionswissenschaft Universität Osnabrück [email protected] Kompaktseminar Universität Tübingen Juli/August 2008 Wednesday Wednesday, 10-12 (s.t) I A choice function approach based on data from St’at’imcets Vera Hohaus: (Matthewson, 1999) I An empirical argument against the choice function approaches – based on VP-ellipsis Ok Soon Jung: (Schwarz, 2004) Wednesday, 13-15.15 (s.t.) I A domain restriction approach towards specificity Christian Eisenreich: (Schwarzschild, 2002) I Wrap up: choice function approaches and domain restriction approaches Kompaktseminar Tübingen 2008 2 I. Choice Function Approaches Choice Functions I An indefinite is interpreted as a free restricted choice function (CF) variable (e.g. Egli and von Heusinger, 1995; Reinhart, 1997; Winter, 1997; Kratzer, 1998; Matthewson, 1999). I A choice function is a function that, when applied to a set, yields an element of this set. A (simplified) choice function definition: (1) [Choice Function 1] CF(fhhe,ti,ei) ↔ ∀Xhe,ti[X 6= ∅ → f (X) ∈ X] I The choice function f introduced by the indefinite is bound via existential closure. Reinhart (1997) and Winter (1997) propose that existential closure is possible from any position. (2) a. Every girl likes some horse. b. ∀x[girl(x) → ∃f [CF (f ) ∧ likes(x, f (horse))]] c. ∃f [CF (f ) ∧ ∀x[girl(x) → likes(x, f (horse))]] Kompaktseminar Tübingen 2008 3 I. Choice Function Approaches Motivation I This approach is motivated by the mentioned problems of the other approaches, in particular the assumption of island free QR. I Even if island-free QR was possible, one would predict exceptional wide scope distributive readings – contrary to fact. (3) a. [three relatives of mine]i IF ti die THEN I will inherit a fortune] b. three rel of mine(x) [die(x) → inherit fortune(I)] I Aim: account for exceptional wide scope and the locality of distributivity. I CF approach inspired by the unselective binding approaches of Heim (1982); Kamp and Reyle (1993), where an indefinite introduces a free individual variable that can be bound at some point. Kompaktseminar Tübingen 2008 4 I. Choice Function Approaches Donald Duck Problem I However, binding of individual variables from some distance delivers wrong results in certain configurations (see Heim 1982; Reinhart 1997): (4) a. If we invite some philosopher, Max will be offended. (Reinhart, 1997) b. ∃x[[philosopher(x) ∧ invite(we, x)] → offended(max)] c. ∃x[philosopher(x) ∧ [invite(we, x) → offended(max)]] I (4a) is represented as (4b), but should really be as (4c). This has been dubbed the Donald Duck problem. I In the CF approach, this problem is non-existent. (5) a. If we invite some philosopher, Max will be offended. b. ∃f [CF (f ) ∧ [invite(we, f (philosopher))] → offended(max)] I Furthermore, the locality of distributivity is predicted. (6) a. If three relatives of mine die I will inherit a fortune. b. ∃f [CF(f ) ∧ [f (three relatives of mine)(die)] → inherit fortune(I)] Kompaktseminar Tübingen 2008 5 I. Choice Function Approaches Intermediate Scope Readings: (Alleged) Advantage I Intermediate scope readings are always derivable (except when ruled out or dispreferred by context). (7) a. Every student has announced that he will leave the party immediately if some lecturer shows up. b. ∃f [CF(f ) ∧ ∀x[student(x) → announce(x, show up(f (lecturer)) → leave party(x))]] c. ∀x[student(x) → announce(x, ∃f [CF(f ) ∧ show up(f (lecturer)) → leave party(x)])] d. ∀x[student(x) → ∃f [CF(f ) ∧ announce(x, show up(f (lecturer)) → leave party(x))]] I Reinhart (1997) and Winter (1997) consider this to be an advantage of their approaches, but Kratzer (1998) thinks it is a disadvantage and makes a different proposal (see below). Kompaktseminar Tübingen 2008 6 I. Choice Function Approaches Intermediate Scope Readings: Problem I Recall that ISRs are sometimes not available even if contextually preferred (contra Reinhart, 1997; Winter, 1997): (8) (Last week, I went to a horse-race every day. It was funny:) # All horses won all races that took place on a day/some day/one day. [∀ horse ∀ race ∃ day] narrow scope contextually excluded [∀ horse ∃ day ∀ race] intermediate scope unavailable [∃ day ∀ horse ∀ race] widest scope contextally excluded Kompaktseminar Tübingen 2008 7 I. Choice Function Approaches The Empty Set Problem I Many people have pointed to the problems of the choice function approaches: (Geurts, 2000; von Stechow, 2000; Ruys, 1999). I Reconsider the definition in (1). I Here comes the empty set problem. (9) a. A green horse ate all the bananas. b. ∃f [CF(f ) ∧ ∀x[bananas(x) → ate(f (green horse), x)]] I Formula (9b) would be true according to Definition 1 under the assumption that no green horses exist and some other element satisfies the sentence predicate ate all the bananas, i.e. (9a) comes out as equivalent to Someone/Something ate all the bananas. Kompaktseminar Tübingen 2008 8 I. Choice Function Approaches How to Overcome the Empty Set Problem I There are different ways out. Geurts (2000): (10) [Choice Function 2] CF(fhhe,ti,ei) ↔ ∀Xhe,ti[[X 6= ∅ → f (X) ∈ X] ∧ [X = ∅ → f (X) = ∗]], where the symbol * denotes a special object which does not satisfy any predicate, i.e. the following holds for every n-ary predicate P : if ai = ∗ for any 1 ≤ i ≤ n, then ha1, . . . , ani is not in the extension of P . I Winter (1997) proposes lifting the type of the choice function: (11) [Choice Function 3] CF(fhhe,ti,hhe,ti,tii) ↔ ∀Xhe,ti[X 6= ∅ → ∃xe[X(x) ∧ f (X) = λPhe,ti.P (x)]] ∧ f (∅he,ti) = ∅hhe,ti,ti Kompaktseminar Tübingen 2008 9 I. Choice Function Approaches The Bound Pronoun Problem I Another problem (Kratzer, 1998; Geurts, 2000): (12) a. Every girl gave a flower to a boy she fancied. b. ∃f [CF(f ) ∧ ∀x[girl(x) → gave flower(x, f (λy[boy(y) ∧ fancied(x, y)]))]] I The derivable and thus predicted wide scope reading in (12b) does not exist. The choice function approach overgenerates. I When there are several girls x who fancy the same group of boys, the argument of f , i.e. λy[boy(y) ∧ fancied(x, y)], is the same for each of these girls. Because f is a function, it cannot yield different results when applied to the same argument. That is, if Miriam and Anne happen to fancy exactly the same group of boys, say Peter, John, and Chris, this representation states that both, Anne and Miriam, must have given a flower to the same boy, e.g. Peter. But this is obviously not a distinct reading for the sentence. Kompaktseminar Tübingen 2008 10 I. Choice Function Approaches How to Overcome the Bound Pronoun Problem I This problem can be overcome in Kratzer’s (1998) approach. Kratzer proposes that CFs should be left free and interpreted by the context. However, CFs can be parameterized. Parameterization is induced by functional dependent elements such as bound pronouns. (See also (Matthewson, 1999) who proposes that CFs can only be closed from the outermost position.) (13) ∀x[girl(x) → gave flower(x, fx(λy[boy(y) ∧ fancied(x, y)]))] I As f is now dependent on the quantifier every girl via the variable x, the choice function can return different results for different values of x, even if the set λy[boy(y) ∧ fancied(x, y)] is the same for these x. Kompaktseminar Tübingen 2008 11 I. Choice Function Approaches Support for Kratzer’s Approach I Kratzer’s and Matthewson’s approaches are supported by the empirical fact that intermediate scope readings are not always available. (14) a. Each teacher overheard the rumor that some student of mine had been called before the dean. b. Each teacheri overheard the rumor that some student of hisi had been called before the dean. (15) a. ∀x[teacher(x) → overheard(x, sent to dean(f (student of mine))] b. ∀x[teacher(x) → overheard(x, sent to dean(fx(student of x))] Kompaktseminar Tübingen 2008 12 I. Choice Function Approaches Further Support for Kratzer’s Approach I There is another advantage of Kratzer’s approach: (16) a. No mani hates a woman hei went to school with. b. ∃f [CF(f ) ∧ ¬∃x[man(x) ∧ hates(x, fx(woman x went to school with))]] I The truth conditions of (16b) are hopelessly weak and (16a) certainly does not have such a weak reading, but all CF approaches predict such a reading, because the representation in (16b) can be derived. man hates • woman man f woman • • −→ • • • −→ • & % • sentence: (16a) false Kompaktseminar Tübingen 2008 representation (16b): true 13 I. Choice Function Approaches Further Support for Kratzer’s Approach I Again, Kratzer’s approach does better. (17) a. No mani hates a/some woman hei went to school with. b. ¬∃x[man(x) ∧ hates(x, fx(woman x went to school with))] I As the function fx gets its value from the context it cannot denote just any function whatsoever. The function must be somehow salient. I And, indeed, (16a) can have a functional wide scope reading: (18) Namely his first girlfriend from school. I However, I doubt that salience is the best way to restrict the domain of functions: (19) a. Every boy brought along some picture of him. b. Strangely enough, it was a picture showing him eating noodle-soup. I A more serious problem: Kratzer’s approach (or better: the choice function mechanism part of her approach) cannot account for all kinds of intermediate scope readings (Chierchia, 2001; Schwarz, 2001). (20) Exactly three students announced to leave the party if some lecturer shows up. Kompaktseminar Tübingen 2008 14 I. Choice Function Approaches A Problem for CF-Kratzer’s Approach I This sentence has the following two intermediate readings: – a functional wide scope reading: there is a function into lecturers such that the number of students such that they announced to leave the party if their functionally corresponding lecturer shows up is exactly three. ⇒ This reading is representable in a CF-framework: (21) ∃f [CF(f ) ∧ |λx.student(x) ∧ announce(x, show(fx(lecturer)) → leave(x))| = 3] – a genuine intermediate scope reading (22) |λx.student(x) ∧ ∃y[lecturer(y) ∧ announce(x, show(y) → leave(x))]| = 3 The number of students such that there is a lecturer and they have announced to leave the party if this lecturer shows up is exactly three. Kompaktseminar Tübingen 2008 15 I. Choice Function Approaches Problems I The readings have different truth conditions: student has announced to leave if . . . shows up lecturer • supervisor • • supervisor • • supervisor • • − −?− → • −−−−−→ −−−−−→ −−−−−→ I The functional wide scope reading is true; the genuine intermediate scope reading is false. ⇒ there is a genuine intermediate scope reading that cannot be represented via the choice function mechanism. I Hence, most CF approaches overgenerate and predict intermediate scope readings that do not exist. Others get rid of this problem, but then undergenerate and cannot account for all kinds of intermediate scope readings. Kompaktseminar Tübingen 2008 16 I. Choice Function Approaches Intermediate Scope Readings vs. Functional Wide Scope Readings I It is important to notice that there is a crucial difference between genuine ISRs and functional wide scope readings: (23) a. Every student will leave the party if some lecturer shows up. b. Every student announced that she will leave the party if some lecturer shows up. I Continuations differ in acceptability: Continuation OK after (23a)? OK after (23b)? Namely, Prof. Humpty (statement of individual) yes yes Her supervisor (statement of functional dependence) yes yes For Ann its Prof. Hob, for Mary Prof. Nob, . . . (pair list) no yes Kompaktseminar Tübingen 2008 17 I. Choice Function Approaches Intermediate Scope Readings vs. Functional Wide Scope Readings I Different continuations ⇒ different kinds of co-variation ⇒ different readings Continuation ∃ lecturer takes. . . statement of individual wide scope [∃ lecturer ∀ student] pair list ISRs [∀ student ∃ lecturer] statement of functional dependence functional wide scope [∃f→lecturer ∀ student] I Function/pair-list distinction occurs with • questions (Groenendijk and Stokhof, 1984; Chierchia, 1993), • functional relative clauses (Sharvit, 1997), • scope phenomena (Endriss, to appear; Ebert and Endriss, 2007). I Note: Functional wide scope readings are no (genuine) intermediate scope readings. Kompaktseminar Tübingen 2008 18 I. Choice Function Approaches Digression: Functional Readings vs. Pair-List Readings I The function/pair-list distinction occurs with scope phenomena (cf. Schwarz, 2001) in the same way as it occurs with e.g. questions (cf. Groenendijk and Stokhof, 1984; Chierchia, 1993) or with functional relative clauses (Sharvit, 1997). I The following example is taken from (Krifka, 2001). (24) Which dish did every guest make? Which dish did most/few guests make? (a) Pasta. (a) Pasta. (b) His favourite dish. (b) Their favourite dish. (c) Al, the pasta; Bill, the salad; . . . (c) # Al, the pasta; Bill, the salad; . . . I Pair-list answers are much more restricted than functional answers. Hence they must be distinguished. Kompaktseminar Tübingen 2008 19 I. Choice Function Approaches Digression: Functional Wide Scope Readings and ISRs are independent I ISRs + functional wide scope readings are independent (cf. Chierchia, 2001; Schwarz, 2001). (25) Exactly two students announced that . . . if some lecturer shows up. I Consider the intermediate scope reading in (a.) vs. the functional wide scope reading (b.): a. The number of students st. there is a lecturer and the students announced that they leave if this lecturer shows up is 2. [exactly two ∃ lecturer IF] b. There is a func. into lect. st. the number of students st. they announced that they leave if the functionally corresponding lect. shows up is 2. [∃f→lecturer exactly two IF] I The readings have different truth conditions: student announced . . . if . . . shows up lecturer John supervisor −−−−−−→ supervisor −−−−−−→ Prof. Humpty − −− → Prof. Hob Ann Mary genuine intermediate scope reading (a.): false Prof. Dumpty functional wide scope reading (b.): true ⇒ ISRs cannot be reduced to functional wide scope readings and vice versa. Kompaktseminar Tübingen 2008 20 I. Choice Function Approaches Digression: Restrictions on Functions I Functional wide scope readings should be dealt with by quantification over functions. I ISRs/narrow scope readings are no functional readings and should be dealt with differently. I But unrestricted quantification over functions leads to problems: • narrow scope representations and functional wide scope representations are often truth- conditionally equivalent (→ Skolemization), e.g. (26) (PL Reading) ∀x[∃y[P (x, y)]] ≡ ∃f [∀x[P (x, f (x))]] (Functional Reading) • We cannot derive two distinct readings to reflect continuation possibilities: (27) a. Every teenager tried to avoid a certain woman at the school party. b. For Peter it was Mary, for Bill it was Paula, etc. ∀x[teenager(x) → ∃y[woman(y) ∧ tried to avoid(x, y))]] c. Namely his mother. ∃f→woman[∀x[teenager(x) → tried to avoid(x, f (x))]] Kompaktseminar Tübingen 2008 21 I. Choice Function Approaches Digression: Restrictions on Functions • Sometimes functional wide scope representations do not correspond to actual readings (cf. Reniers, 1997; Schwarz, 2001; Chierchia, 2001). (28) No X V some Y . Consider all situations where each X does not V all Y s, e.g. • Then there is always a function f that verifies the sentence: (29) ∃f [¬∃x[V (x, f (x))]] Kompaktseminar Tübingen 2008 ≡ ∀x[∃y[¬V (x, y)]] 22 I. Choice Function Approaches Digression: Restrictions on Functions I Hence, we have to restrict the stock of functions that can plausibly be quantified over. I In the literature, it has been proposed that only ‘natural’ (Chierchia, 1993; Jacobson, 1999), ‘salient’ (Sharvit, 1997; Kratzer, 1998, 2003) or ’nameable and informative’ (Endriss, to appear) functions are involved in the case of functional readings. I Crucially, an enumeration of ordered pairs is not regarded as ‘natural’, ’salient’, or ’nameable and informative’. I Restriction to natural functions distinguishes the functional wide scope reading from the narrow scope reading: (30) ∃f→woman[f is natural ∧ [∀x[teenager(x) → tried to avoid(x, f (x))]]] and the derivation of unavailable readings is suppressed. (31) ∃f [f is natural ∧ [¬∃x[V (x, f (x))]]] Kompaktseminar Tübingen 2008 23 I. Choice Function Approaches Digression: Functional Wide Scope Readings are no ISRs I Let’s go back to the contrast between (23a) and (23b). a. Every student will leave the party if some lecturer shows up. b. Every student announced that she will leave the party if some lecturer shows up. I Recall: (23b) has a genuine ISR, while (23a) has not. Continuation Reading Namely, Prof. Humpty wide scope [∃ lecturer ∀ student] For Ann its Prof. Hob, . . . intermediate scope [∀ student ∃ lecturer] His supervisor functional wide scope [∃f→lecturer ∀ student] I Note: Often people do not distinguish between ISRs and functional wide scope readings. Only the availability of ISRs, but not of functional wide scope readings, is restricted. Kompaktseminar Tübingen 2008 24 II. Domain Restriction Approaches Extreme Domain Restriction I The restrictor set of the indefinite is implicitly domain restricted. I The speaker (or someone else) does not have a referent, but has a certain property in mind. I So, here again, there is a speaker-hearer asymmetry. I This property can be so specific that it holds for only one individual. It denotes a singleton set (e.g. Schwarzschild, 2002). (32) a. If a friend of Dena comes to the party, Maria will leave immediately. b. ∃x[friend of Denawith certain property (x) ∧ comes to party(x)] → leave(Maria) I The indefinite is interpreted in situ and yet receives an apparent wide scope interpretation. Furthermore, domain restriction is a mechanism we need anyway. (33) a. Yesterday, when Maria came to school she saw that every teacher was wearing red trousers. b. every teacher 6≡ every teacher in the world c. every teacher ≡ every teacher in Maria’s school/every teacher Maria met yesterday Kompaktseminar Tübingen 2008 25 II. Domain Restriction Approaches Extreme Domain Restriction I Going back to the problematic examples from before: (34) a. I bet that, after the next Olympic Games also, everyone will believe that some athlete had taken illegal substances. Namely the one that will have won most medals. b. (Before the 100m race has actually started, someone says:) All the journalists will try to get an interview with ONE athlete/a certain athlete after the race. Namely the winner of the race/the one that will have won. Kompaktseminar Tübingen 2008 26 II. Domain Restriction Approaches No Property in Mind I One can construe examples of exceptional wide scope readings where the speaker neither knows the intended referent nor a defining property of this referent (nor does anyone else). (35) a. Imagine a group of people playing a (very strange) game. The rules are as follows: if a certain word that is agreed upon by the group is uttered by one person, everyone has to touch the floor with their hands. Furthermore, if another word that has been agreed on is mentioned, everyone has to clap their hands. Then Maria enters the scene and asks about the game. Max explains the rules. Then Maria asks back: b. (So we are playing several runs of the game, right? And for every run the following holds:) If some word that we agree on beforehand is uttered, we all have to touch the floor as quickly as possible. c. (Also es gibt verschiedene Runden ja?) Und in jeder Runde ist es so: wenn ein Wort, das vorher ausgemacht wurde, fällt, müssen wir alle so schnell wie möglich den Boden berühren. I Domain restriction approaches cannot account for the salient reading, indefinite seems to take genuine wide scope. Kompaktseminar Tübingen 2008 27 II. Domain Restriction Approaches Partitives I It is not always plausible to assume implicit domain restriction to a singleton set, e.g. with partitives. (36) If one of my relatives dies, I will inherit a fortune I An exceptional wide scope reading is available. I But partitives usually indicate that the restrictor set contains more elements than the ones just talked about. (37) a. One of my Mexican relatives rang me up, today. b. A Mexican relative of mine rang me up, today. I The partitive construction of (37a), but not the sentence with the ordinary indefinite in (37b) implies that I have more than one relative. Kompaktseminar Tübingen 2008 28 Bibliography Gennaro Chierchia. A Puzzle About Indefinites. In Carlo Cecchetto, Gennaro Chierchia, and Maria Teresa Guasti, editors, Semantic Interfaces. CSLI Publications, Stanford, 2001. Gennaro Chierchia. Questions with Quantifiers. Natural Language Semantics, 1:181–234, 1993. Christian Ebert and Cornelia Endriss. Functional Topics. In Estela Puig-Waldmüller, editor, Proceedings of the conference ‘sub 11 – Sinn und Bedeutung’. 11th annual meeting of the Gesellschaft für Semantik, pages 194–208. Universitat Pompeu Fabra, Barcelona, 2007. Urs Egli and Klaus von Heusinger. The Epsilon Operator and E-Type Pronouns. In Urs Egli, Peter E. Pause, Christoph Schwarze, Arnim von Stechow, and Götz Wienold, editors, Lexical Knowledge in the Organization of Language, number 114 in Current Issues in Linguistic Theory, pages 121–141, Amsterdam, 1995. Benjamins. Cornelia Endriss. Quantificational Topics. A Scopal Treatment of Exceptional Wide Scope Phenomena. Studies in Linguistics and Philosophy. Springer, Dordrecht, to appear. Master Thesis. Bart Geurts. Indefinites and Choice Functions. Linguistic Inquiry, 31:731–738, 2000. Jeroen Groenendijk and Martin Stokhof. Studies on the Semantics of Questions and the Pragmatics of Answers. PhD thesis, University of Amsterdam, Amsterdam, 1984. Irene Heim. The Semantics of Definite and Indefinite Noun Phrases. PhD thesis, University of Massachusetts, Amherst, 1982. Pauline Jacobson. Towards a Variable-Free Semantics. Linguistics & Philosophy, 22(2):117–184, 1999. Hans Kamp and Uwe Reyle. From Discourse to Logic. Kluwer Academic Publishers, Dordrecht, 1993. Angelika Kratzer. A Note on Choice Functions in Context. Ms, 2003. Angelika Kratzer. Scope or Pseudoscope? Are there Wide-Scope Indefinites? In Susan Rothstein, editor, Events and Grammar, pages 163–196. Kluwer Academic Publishers, Dordrecht, 1998. Manfred Krifka. Quantifiying Into Question Acts. Natural Language Semantics, 9:1–40, 2001. Lisa Matthewson. On the Interpretation of Wide-scope Indefinites. Natural Language Semantics, 7:79–134, 1999. Tanya Reinhart. Quantifier Scope: How Labour is divided between QR and Choice Functions. Linguistics & Philosophy, 20: 335–397, 1997. Kompaktseminar Tübingen 2008 29 Fabien Reniers. How to (S)cope with Indefinites. Master’s thesis, University of Utrecht, 1997. Eddy Ruys. Unexpected Wide Scope Phenomena. Ms, 1999. Bernhard Schwarz. Two Kinds of Long-Distance Indefinites. Manuscript, 2001. Bernhard Schwarz. Indefinites in Verb Phrase Ellipsis. Linguistic Inquiry, 35(2):344–353, 2004. Roger Schwarzschild. Singleton Indefinites. Journal of Semantics, 19.3:289–314, 2002. Yael Sharvit. The Syntax and Semantics of Functional Relative Clauses. PhD thesis, Rutgers University, 1997. Arnim von Stechow. Some Remarks on Choice Functions and LF-Movement. In Klaus von Heusinger and Urs Egli, editors, Reference and Anaphoric Relations, pages 193–228. Kluwer Academic Pubishers, Dordrecht, 2000. Yoad Winter. Choice Functions and the Scopal Semantics of Indefinites. Linguistics & Philosophy, 20:399–467, 1997. Kompaktseminar Tübingen 2008 30
© Copyright 2026 Paperzz