意味論研究会2017年6月23日

(Joint work with Yasutada Sudo.)

The existence of anti-rogative predicates, i.e., clause-embedding predicates that only select for declarative complements (e.g., “hope”, “believe”, “think”), poses a problem for the semantics of clausal embedding. If embedded questions can be reduced to propositions under some predicates such as “know (by e.g., the answerhood operator; Heim 1994, Dayal 1996), it is not clear why the same mechanism cannot apply to anti-rogative predicates. On the other hand, if a semantic type distinction is made between anti-rogative and question-embedding predicates, it remains unclear why the distinction correlates with certain lexical semantic properties, such as neg-raising property (Zuber 1982; Thriller et al. 2016) and factivity (Egré 2008).

In this talk, we point out another cross-linguistically stable correlation between anti-rogativity and lexical semantics, namely that all non-factive preferential predicates, such as “wish” and “hope”, are anti-rogative, and provide a semantic analysis of the correlation. We show that this correlation can be straightforwardly explained in a framework such as Inquisitive Semantics in which complements uniformly denote sets of propositions (Uegaki 2015; Theiler et al. 2016), when coupled with independently motivated semantics of factivity and preferentiality. In particular, we show that the ordering-based semantics for preferential predicates (e.g., Heim 1992; Villalta 2008) leads to systematic triviality when the whole set of propositions denoted by an interrogative complement enter into the ordering relation, and that this happens precisely when the predicate is non-factive preferential, as in the case of “wish” and “hope”.