Which judgments show weak exhaustivity? (And which don’t?)

Abstract

This paper considers two of the most prominent kinds of evidence that have been used to argue that certain embedded questions receive weakly exhaustive interpretations. The first kind is exemplified by judgments of consistency for declarative sentences that attribute knowledge of a wh-question and ignorance of the negation of that question to the same person, and the second concerns asymmetries between the role of positive and negative information in validating question-embedding surprise ascriptions, and similar judgments for other attitudes. I argue that neither type suffices to show weak exhaustivity. The first can be analyzed in terms of strong exhaustivity in combination with domain restriction effects, while the second can be analyzed in terms of a mention-some interpretation. These kinds of evidence have served as the empirical basis for many claims about weakly exhaustive readings, so the observation that they are unreliable calls into questions a large body of established work on the semantics of question embedding.