意味論研究会2013年6月28日

It has been observed that the exhaustive inference (hereafter ExhInf) of question-answers arises only when the polarity of the answer matches that of the question (Schulz & van Rooij 2006, Spector 2007). It is claimed that, e.g., although the answer "I will invite Sue" to the question "Who will you invite?" gives rise to the inference that Sue is the only person that I will invite, the answer "I won’t invite Sue" to the same question does not readily give rise to the inference that Sue is the only person that I will not invite (pace Zimmermann & von Stechow 1984).

Previous approaches to this phenomenon stipulate mechanisms specific to the polarity or monotonicity-mismatching question-answer pairs (Schulz & van Rooij 2006, Spector 2007). In this talk, I provide an account of the phenomenon in terms of a general constraint on the alternative utterances to be used in the generation of ExhInf. In a nutshell, the constraint states that an alternative has to be congruent to some immediate Question under discussion (IQUD; Roberts 1996), given an extended notion of Question-Answer Congruence (Rooth 1992). This constraint restricts the alternatives for a positive answer to be only positive alternatives, predicting the ExhInf. On the other hand, the negative answer is licensed only when an additional QUD is accommodated in the context, resulting in a larger set of alternatives including both positive and negative propositions, which in turn leads to the absence of ExhInf (Sauerland 2004). In the talk, I will discuss further predictions of the current approach, and discuss its differences from existing analyses.