Semantic Constraints on Interrogative Embedding

• Applicant: Professor Dr. Clemens Steiner-Mayr
• Subject Area: General and Comparative Linguistics, Experimental Linguistics, Typology, Non-European Languages
• Term: from 2016 to 2021
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 314523601

Most languages embed both declarative and interrogative clauses. For example, English 'know' and 'believe' in 'John knows / believes that Mary smokes' embed the declarative 'that Mary smokes'. But only 'know' in 'John knows / *believes who smokes' can embed the interrogative 'who smokes'. In theoretical linguistics, Grimshaw (1979, Linguistic Inquiry) accounted for such data by postulating 's-selection'. This is a lexical specification of predicates as to what type of clause they combine with. This view has been widely adopted and basically stands till today. Adger & Quer (2001, Language), however, point out a difficulty for this approach (cf. Eckardt 2007, in On Information Structure, Meaning and Form). For instance, 'be certain' only combines with an interrogative when it is negated: 'John isn't / *is certain who smokes'. But it is not obvious how the linguistic context can influence the lexically determined s-selection property of 'be certain'. In response to this problem, this project explores a novel way of viewing embedded interrogatives: I investigate whether they are similar to negative polarity items like any and whether their distribution is therefore similarly constrained. Specifically, a compositional semantics based on a grammatical notion of pragmatics (Fox 2007, in Presupposition and Implicature in Compositional Semantics, Chierchia 2013, Logic in Grammar) distinguishes between the cases where `be certain' cannot embed interrogatives and those cases where it can: a contradiction and thus unacceptability arises without negation, but not so with negation. This offers a new perspective on interrogative embedding. It is no longer predicates alone which can be said to allow for or to resist interrogative embedding, but rather, it is whole linguistic contexts. Specifically, the combination of the lexical semantics of a predicate and the so-called polarity property of the sentence it appears in both affect interrogative embedding. As a consequence s-selection can be eliminated. This view accounts for many facts that have so far resisted a principled explanation. In order to achieve this, the project carefully investigates how the lexical semantics of predicates affects interrogative embedding. For instance, the factivity of `know' ensures that interrogatives can be embedded regardless of polarity. The neg-raising property of `believe', however, does the exact opposite. The project furthermore explores linguistic contexts beyond negation and how they affect the embedding of interrogatives. This is shown to, for instance, cast light on the hitherto puzzling interpretive properties of communication verbs like tell when embedding interrogatives. In order to critically evaluate the suggested system, a cross-linguistic investigation of interrogative embedding is part of the project. Combining these lines of investigation, the project culminates in a novel unified semantics for declarative and interrogative clause embedding.

• DFG Programme: Research Grants