Assertion and Presupposition: the Syntax of Embedded Context Updates

This paper investigates the idea that the presence vs. absence of illocutionary potential in declarative clausal complements is syntactically reflected in the embedded clause. Specifically, we examine the two claims that (a) clauses with illocutionary potential involve an extended left-periphery encoding elements relevant to assertion (e.g. Rizzi 1997), and (b) that clauses without illocutionary potential are referential, presuppositional, or given, encoding syntactic properties of definiteness (Kiparsky & Kiparsky 1970; et seq). To this aim, we look at three phenomena commonly used to distinguish assertive and definite clauses (the licensing of Main Clause Phenomena, clausal anaphora, and extraction), asking whether their distribution across embedding contexts in fact tracks this pragmatic distinction. By looking at the distribution of these phenomena across a fairly wide range of predicates of different types (including both factive and non-factive verbs, as well as both negative and positive predicates), we conclude that the phenomena in question neither share the same distribution, nor are they sensitive to the same properties of the (embedding) context. Thus, they do in fact not call for a unified account. Rather, we find that among the phenomena investigated, only one, namely embedded V2, fits the description above of being licensed in contexts with illocutionary potential, and not available in contexts where the embedded proposition is obligatorily discourse old, or `pragmatically definite'. We conclude the paper with a proposal for a formal characterization of the types of embedded context updates that license embedded V2; extending the so-called Table Model of Bruce & Farkas 2010 to complex speaker assertions. We further show how this semantic model can be straightforwardly implemented in a standard (Rizzian) ForceP-based analysis of the assertive left-periphery. This account thus provides a concrete semantics for complex assertions, and makes explicit the bridge between this type of (fairly standard) syntactic approach to embedded assertions/V2, and the pragmatic approach to embedded V2 proposed in Caplan & Djärv 2017 and Djärv 2019.