Semantics of nominal and clausal embedding: how (not) to embed a clause and why

Abstract

There is a large class of verbs in English which can embed either a nominal or a clause.

(1) a. Copernicus announced/believes/clarified/discovered/explained [CP that the earth revolves around the sun]. b. Copernicus announced/believes/clarified/discovered/explained [DP the theory]. These clause embedding verbs (CEVs) have been a focus for several strands of recent and not-so-recent work in both Linguistics and Philosophy. In Linguistics these verbs have been of interest to theories of argument selection (Grimshaw (1990), Pesetsky (1996)) and semantic composition (Kratzer (2006)), since CPs are non-prototypical arguments from either a semantic or syntactic perspective. These verbs have also long been important to Philosophers of Language interested in the role that ‘propositions’, which under the standard account are the denotation of these ‘that-clause’ (TC) CPs, play in the semantics of ‘propositional attitude reports’ and related modal and intensional constructions (Prior (1971), King (2002), Moltmann (2003)).

This thesis argues for a novel account of the compositional semantics of CEVs which takes TCs to denote properties of contentful individuals that have two pathways to combine with a CEV, either through restriction or saturation of an internal argument. This account builds heavily on the Predicativist proposal of (Kratzer (2006), Moulton (2009, 2015)) which treats TCs as semantically predicates, in contrast to the standard view in which they denote propositions. Crucially for such an account TCs are not treated as thematic arguments of CEVs, but as modifiers of their objects. I argue that this theory is fundamentally correct, but that empirical observations about the behaviour of ‘presuppositional’ CEVs with respect to their available substitutions (Bach (1997)) and entailments (Uegaki (2015)) demonstrate the need to modify the theory further.

The modification that I suggest exploits a recent, independent syntactic argument which demonstrates that some TC complements to CEVs are not bare CPs, but CPs headed by a covert determiner (Kastner 2015). I argue that augmenting the Predicativist semantic proposal with this syntactic claim, along with standard compositional tools allows us to explain a variety of data which was puzzling under pre-existing theories. The presuppositional DPs that result from combining a covert determiner with a predicative clause (denoting some definite individual with the proposition inside the embedded clause as its content) compose with the CEV by saturating its internal argument position, unlike the bare CPs which combine by restricting it. The new proposal has the advantage of capturing the classic puzzles for CEVs, including the entailment failures which were difficult for the unaugmented Predicativist account, whilst also being the result of a natural extension of standard compositional semantics for definite DPs with embedded nominas.

This approach also provides novel insights into generalisations from the linguistic literature on argument selection and case, as it makes concrete predictions about how clausal ‘arguments’ interact with the case and theta systems. These predictions can be tested against a number of classically difficult phenomenon like experiencer predicates (Reinhart (2003)) and the verb ‘explain’ (Pietroski (2000)). I demonstrate how these predictions are borne out, supporting the proposal and advancing our understanding of these topics. The thesis is structured as follows: Chapter 1 presents a series of puzzles derived from TC and DP embedding under CEVs, namely, Quantificational inferences, Fine-grained semantic selectional restrictions, the variability of attitudinal objects, and content noun entailment patterns. I argue that no existing account is capable of explaining all of these puzzles, and that the dominant Propositionalist theory must be mistaken in several of its key assumptions. In Chapter 2 I introduce the alternative Predicativist account of the semantics of TCs (Kratzer (2006), Moulton (2009, 2015)), discuss the motivations for such an account, and I provide a novel argument for the position based on copula constructions with post-copula TCs. I then show how this theory accounts for most of the puzzles introduced in chapter 1, but cannot deal with the entailment failures of a certain subclass of CEVs. Chapter 3 then suggests that this class of entailment failing verbs is coextensive with the class of verbs identified in Kastner (2015) as selecting for TCs which are syntactically DPs and interpreted as presuppositional. I provide evidence for this analysis and discuss some cross-linguistic analogues with overt determiners with CP complements. I then propose a compositional semantics for these presuppositional TCs adapted from the Predicativist account argued for in chapter 2. I then show how the resultant account resolves all of the empirical puzzles set up in chapter 1, without introducing any new syntactic or semantic machinery that has not already been independently proposed. Chapter 4 explores the predictions and consequences of this theory for other domains of the selection literature, in particular relating to ECM, factivity, question-embedding, communicatives, and experiencer alternations. I argue that the proposal meshes well with existing proposals about how to understand case-assignment and argumenthood (Stowell (1982)). In the conclusion I discuss some consequences for the Philosophical literature on propositional attitudes, and present cross-linguistic and historical evidence for the plausibility of my account which invite future empirical work on the topic