Ezra Keshet will give a guest presentation in the Topics in Syntax seminar on Thursday. The talk will begin at 4:00 PM and will be in Machmer W-27 (not the room the seminar is normally scheduled for).
Title: Answering Questions about Coherence and Anaphora"
Abstract:
As I will argue in my general-audience talk (and preview here), discourse coherence relations between sentences can constrain pronoun referents in a way suitable to generate sloppy readings in ellipsis contexts. For instance, sentence (1d), which stands in a RESULT relation to (1c), most easily means that people didn’t vote for Nixon:
(1) a. Kennedy looked good on TV. b. People voted for him.
c. Nixon looked bad on TV. d. People didn’t.
In this talk, I claim that coherence relations are a type of projective meaning (see Tonhauser et al. 2013) that determines a hidden Question Under Discussion (QUD) as defined by Roberts (2012). For instance, (1b) is understood as a response to the (hidden) QUD What resulted from Kennedy looking good on TV? while (1d) is the response to QUD What resulted from Nixon looking bad on TV? (See Kehler 2009) for a previous suggestion along these lines.)
Pronouns in the answers to questions are often quite constrained in their interpretations:
(2) Q: Who does Bill like? A: He likes Mary. [He must refer to Bill]
I argue that these constraints on pronouns follow from a very small modification to theories of Focus/Givenness such as Rooth (1992) and Schwarzschild (1999). Under Rooth’s system, for instance, a ~ (‘squiggle’) operator effectively ensures that the answer in (2) appears in the Hamblin (1973) denotation of the question. The ~ operator, I argue, can alter the local assignment function to achieve its purpose, setting the referents of pronouns such as he in (2) to properly fit them into the question denotation. With this machinery in place, the pronoun him in (1b) can receive a locally-derived meaning (Kennedy) via the ~ operator but this meaning can have a different extension (Nixon) when picked up by the ellipsis site in (1d).
Next, I extend this analysis to embedded coherence relations, especially in quantified sentences containing a donkey pronoun, such as (3). The “consequent” clause he cries in (3) stands in a RESULT relation to the “antecedent” clause Jill teases a boy. In fact, this consequent stands in several different RESULT relations to the antecedent – one for each situation quantified over by whenever. I argue that the consequent clause is therefore evaluated with respect to a slightly different QUD for each situation quantified over, analogous to the way the phrase his sister in (4) generates a slightly different presupposition for each boy quantified over by every. Since each QUD for the consequent potentially pertains to a different boy, the pronoun he in he cries can end up referring to different boys (but in each situation refers to the boy that Jill teased in that situation).
(3) Whenever Jill teases a boy, he cries.
(4) Every boy loves his sister.
