Ezra Keshet (University of Michigan) will present "Two Effects of Context on Truth-Conditional Meaning" at the department colloquium this Friday (February 8) at 3:30PM in Machmer E-37. An abstract of his talk follows.
The lexical items and syntactic structure of a sentence play a large role in its interpretation, but other factors often contribute as well, such as the time/place of utterance, other sentences in the same discourse, salient surrounding objects; extralinguistic goals; and more. This talk presents case studies highlighting the semantic effects of two such suprasentential factors: focus/information structure, and discourse coherence.
The first study involves so-called conditional conjunctions such as (1) whose meanings are quite similar to conditional sentences involving the word if as in (2):
(1) You get there early enough, and you usually find a seat.
(2) If you get there early enough, you usually find a seat.
I will present evidence that such sentences require a particular information structure, namely one where the first clause is Given in the context (or accommodated as Given). I argue that the conditional meaning results from the interaction between a modal quantifier such as usually in (1) and this information structure. A standard if-clause generally serves to restrict a modal quantifier (see Kratzer 1986) and since Given material tends to join the restriction of quantifier as well, this information structure mimics the effect of an if-clause.
The second study examines how discourse coherence can constrain pronoun referents in a way previously thought to require syntactic c-command. Coherence is the name for the unspoken links between clauses in discourse, as shown in (3), which is most easily understood to mean (4):
(3) Ezra flew into Hartford. He’s giving a talk at UMass.
(4) Ezra flew into Hartford because he’s giving a talk at UMass.
Hobbs (1979) noted that the establishment of such coherence relations also constrains pronoun reference, for instance constraining he in (3) to refer to Ezra when the discourse is understood as in (4). Ross (1967) described two readings for pronouns in ellipsis, as shown in (5). Most linguists, following Reinhart (1983), claim that the sloppy reading requires a c-commanding antecedent, such as Bill, which c-commands his in (5). I argue instead that discourse coherence constraints on pronouns can also generate sloppy readings, as shown in (6) and (7).
(5) Bill loves his dog. John does, too.
a. ...John loves Bill’s dog, too. (strict reading)
b. ...John loves John’s dog, too. (sloppy reading)
(6) The man who called Jane asked her out. The man who emailed Vera did ask Vera out, too.
(7) Kennedy looked good on TV. People voted for him.
Nixon looked bad on TV. People didn’t vote for Nixon