The Speech Report: A Cross-Linguistic Investigation into the Compositional Semantics of Reported Speech

Abstract

This dissertation investigates how the morphosyntax of embedded clauses determines their interpretation in speech and attitude reports. The central question is whether there is a one-to-one mapping between the syntactic status of a complement clause and the reading it receives. Nominal, argument-saturating clauses are predicted to give rise exclusively to Content of Theme readings, where the clause describes the theme of the matrix predicate; non-nominal, predicate-modifying clauses are predicted to give rise exclusively to Content of Event readings, where the clause describes the content of the matrix event. I test this mapping against novel cross-linguistic data from five typologically diverse and genetically unrelated languages — Catalan, Georgian, Kazakh, Korean, and Mandarin — collected through a large-scale survey designed to separate the morphosyntactic properties of clauses from their interpretations in context. I first investigate the claim that argument clauses are exclusively interpreted as Content of Theme (Özyıldız and Uegaki, 2024). I find numerous apparent Content of Event readings for nominal clauses in Georgian, Kazakh, and Korean, and argue that these are False Positives: the embedded clause describes the theme, but the predicate’s semantic requirements happen to be compatible with a context in which the theme’s content matches what the speaker says. Whether a False Positive arises depends on two dimensions: the denotation of the embedded clause, whether it describes a situation that exemplifies a proposition (a Sit-CP, following Bondarenko 2022) or an entity whose content is that proposition (a Cont-CP), and the predicate’s restriction on the relation between the content of its event argument and its theme. I argue that degen-headed clauses are ambiguous between a grammaticalized reading and a decomposed reading in which degen is the past participle of dew “say”. Under the decomposed reading, the clause compositionally describes a speech event, which restricts the contexts in which False Positives arise and explains why these clauses pattern differently from other argument clauses under predicates of explanation and commentary. I then test whether the Missing Answer Reading— the unavailability of propositional readings for predicate- modifying interrogatives (Özyıldız, 2019b; Özyıldız and Uegaki, 2024) — holds across the five languages. The generalization survives intact. Apparent counterexamples in Catalan, Georgian, Korean, and Mandarin each turn out to involve a clause that is ambiguous between interrogative and declarative mood, or ambiguous between argument and predicate-modifier status. Finally, I argue that propositional readings of embedded interrogatives arise from clause-internal answer- hood operators rather than from predicate-internal mechanisms, contrary to Question-Oriented accounts (Ue- gaki, 2023; Özyıldız and Uegaki, 2024). I defend a typology of three interrogative denotations: entities with question content, entities with propositional content derived via an answerhood operator, and exemplifying situations derived from the question’s answer. I argue that Kazakh non-finite participles instantiate the third kind, extending Bondarenko (2022)’s declarative typology to interrogatives.