Donkey sentences have existential and universal readings, but they are not often perceived as ambiguous. I extend the pragmatic theory of homogeneity in plural definites by Kriz ( 2016 ) to explain how context disambiguates donkey sentences. I propose that a semantic theory produces truth value gaps in certain scenarios, and a pragmatic theory fills these gaps in context-dependent ways. By locating the parallel between donkey pronouns and definite plurals is located in the pragmatics rather than in the semantics, I avoid problems known to arise for some previous accounts according to which donkey pronouns and definite plurals both have plural referents ( Krifka 1996 ; Yoon 1996 ). I sketch an extension of plural compositional DRT ( Brasoveanu 2008 ) that delivers the required truth value gaps by building on concepts from error-state semantics and supervaluation quantifiers. Proceedings of SALT 26: 684–704, 2016 Homogeneity in donkey sentences * Lucas Champollion New York University Abstract Donkey sentences have existential and universal readings, but they are not often perceived as ambiguous. I extend the pragmatic theory of homogeneity in plural definites by Križ (2016) to explain how context disambiguates donkey sentences. I propose a semantic theory that produces truth value gaps in certain scenarios, and a pragmatic theory that fills these gaps in context-dependent ways. By locating the parallel between donkey pronouns and definite plurals in the pragmatics rather than in the semantics, I avoid problems known to arise for some previous accounts according to which donkey pronouns and definite plurals both have plural referents (Krifka 1996; Yoon 1996). I sketch an extension of plural compositional DRT (Brasoveanu 2008) that delivers the required truth value gaps by building on concepts from error-state semantics and supervaluation quantifiers. Keywords: donkey sentences, trivalence, weak/strong (existential/universal) ambiguity, extension gaps, pragmatics 1 Introduction It is an old observation that some donkey pronouns seem to be understood as having existential force and others as having universal force. The following pair is adapted from Yoon 1996: (1) Usually, if a man has a garage with a window, . . . a. he keeps it open while he is away. b. he keeps it closed while he is away. On the most plausible reading of (1a), the donkey pronoun it could be paraphrased as one of the windows in his garage (except that there is no implication that the * I am indebted to Robert Henderson for extensive discussions of the contents of this paper and for specific and very helpful suggestions regarding the implementation of error state semantics and supervaluation quantifiers in PCDRT. I am grateful to Chris Barker, Justin Bledin, Adrian Brasoveanu, Dylan Bumford, Jan van Eijck, Makoto Kanazawa, Manuel Križ, Sophia Malamud, Philippe Schlenker, Anna Szabolcsi, the NYU semantics group, and the audience at SALT for helpful feedback. All errors are mine. ©2016 Champollion Champollion Let me now turn to the restrictor. Brasoveanu (2008) proposes that indefinites are ambiguous between a maximal or “strong” and a nonmaximal or “weak” inter- pretation. Maximal indefinites simultaneously introduce as many values as possible, while nonmaximal indefinites are free to assign a smaller set. For example, the as- signments in any output state of ad donkey in (19) map d to farmer-owned donkeys. If ad is maximal, these assignments do this in such a way that no farmer-owned donkey is left out. If ad is nonmaximal, among the output states of the indefinite there will be some whose assignments leave out some donkeys. In Brasoveanu 2008, the main purpose of this ambiguity is to account for the∃/∀ ambiguity. But this is no longer needed on the present theory. For this reason, I assume that indefinites always receive a maximal interpretation. Brasoveanu (2010), Homogeneity in donkey sentences sets. The “inner envelope” contains every donkey-owning farmer who beats all of his donkeys. The “outer envelope” contains every donkey-owning farmer who beats at least one of his donkeys. A supervaluation quantifier succeeds if the underlying ordinary quantifier is true of all sets that are supersets of the inner envelope and subsets of the outer envelope; it fails if the underlying ordinary quantifier is false of all these sets; otherwise, it produces an error. This is implemented in (34) and (35). Some auxiliary definitions follow. The definitions in (36) are motivated and explained in Brasoveanu 2010. Definition (37) specifies the sets that are sandwiched between the inner envelope f Iε=⋆ and the outer envelope f I associated with f . The entry (34), slightly modified from Brasoveanu 2010, first stores under f all those individuals that satisfy the restrictor and then copies from f into f′ all those individuals that also satisfy the nuclear scope. Finally, f and f′ are passed to the supervaluationist determiner, defined as in in (35). Here, the input state is passed on unchanged if the underlying determiner accepts all precisifications of the nuclear scope; if the determiner accepts some but not all of these precisifications, an error is generated. This makes the determiner behave supervaluationistically. (34) det f⇝ λPλP′.max f (⟨ f⟩P( f ));max f′ ([ f′⊑ f ];⟨ f′⟩P′( f′)); DETsv{ f , f′} (35) Let DET be a static determiner. Then DET sv{ f , f′} def ={S| f Iε=⋆⊆ S⊆ f I} 7 Conclusion Definite plurals and donkey sentences can be given a uniform pragmatic treatment. Previous accounts relied on the problematic assumption that it and the donkey(s) he owns can be given a parallel analysis in terms of plural individuals. However, Kanazawa (2001) showed that plural individuals cannot be involved in the semantics of it. This paper avoids the need for plural individuals by leveraging PCDRT’s evaluation-level pluralities. Aided by error-state and supervaluation semantics, PCDRT delivers trivalent semantics to the pragmatics in a fully compositional way. 701