Colloquium: Towards a theory of processing Question-Answer relations

April 14, 3:30-5pm, Cobb 201
Lyn Frazier, University of Massachusetts, Amherst 

There has been a surprising lack of psycholinguistic research on the processing of question-answer pairs. This is striking because asking and answering questions is such a basic use of human language. Further, it may turn out that implicit questions and answers, or comments, may underlie the structuring of discourse if the Question-under-Discussion (QUD) approach is correct. In this talk I will discuss a number of acceptability judgment and processing studies of question-answer pairs. The first half of the talk will focus on the relation required for answers to be fully grammatical. It will be argued that an answer to a question is most acceptable when it matches the syntactic form of the question. With ellipsis, an answer is fully grammatical only if the overt material in the answer has moved to the specifier of a focus projection, as argued by Merchant (2004). In an example like (1a) below, a direct fragment answer is permissible because a that-clause may front (e.g., That he lied, Sam admitted); by contrast, without that (1b), a fragment answer analysis is not fully grammatical (e.g., *He lied, Sam admitted.). When no that is present, a sentence following a question (1b) may be interpreted as an indirect reply, though sometimes this will be odd, depending on the content of the sentence.

An overt question sets up a QUD and establishes an expectation concerning the Source (e.g., in (1a) the person responsible for the embedded proposition, namely, Sam). A direct fragment answer will resolve the question/QUD and confirm the expectation concerning the Source, as in (1a).

(1)a.Speaker A: What did Sam say?
  Speaker B: That he lied. (Direct fragment answer: Sam is the Source for p; p=Sam lied.)
 b.Speaker A: What did Sam say?
  Speaker B’: He lied.     (Indirect reply: B’ is the source for p; p =Sam lied.)

An indirect reply will often disconfirm the expectation concerning the Source, as in (1b). A restoration study will be presented showing that, when presented masked by noise, the complementizer that is restored more often when the Source of the answer differs for the direct answer (1a) and the indirect reply analysis (1b) than for examples where the two analyses share the same Source (Clifton, Frazier, Harris, Mack, in progress).

Does an implicit QUD influence sentence processing? In prior work, Frazier and Clifton (2005) argued for the Main Assertion principle, which specifies that elided constituents preferably find their antecedents in the main assertion of the preceding utterance. I will argue that the Main Assertion principle is not sufficiently general and it should be replaced by a general interpretation principle favoring analyses that comment on the QUD. Finally, evidence will be presented showing that a question influences the processing of an answer during the online processing of the answer, as expected on the view where questions and QUDs are the basic mechanism for structuring a discourse.