In this thesis, I propose unifying unembedded and embedded rising declaratives by adopting an approach that treats them as the same phenomenon. First, I present evidence that embedded rising declaratives should be united with their unembedded counterparts. Both embedded and unembedded rising declaratives share an intonation, as well as discourse effects. As part of unifying them, I develop a taxonomy of embedded and unembedded rising declaratives. This taxonomy breaks them into two types - one that seems to raise a question (called an inquisitive rising declarative), and one that seems to make an assertion (called an assertive rising declarative). I then explore what can and cannot embed a rising declarative, with a discussion of Simons (2007) and the predicates she discusses. Second, I describe two previous approaches that can account for either unembedded inquisitive rising declaratives or unembedded assertive rising declaratives (Farkas & Roelofsen, 2017; Malamud & Stephenson, 2015). I also discuss two approaches that do treat unembedded inquisitive rising declaratives and assertive rising declaratives as united phenomena (Goodhue, 2021; Jeong, 2018). Then, I build out a theory of rising declaratives where rising declaratives are a discourse move that speakers can make when they are attempting to resolve the main issue through a series of questions, called a Strategy. As part of this, the rise in rising declaratives contributes unsettledness about one of the questions in the Strategy. I propose a way of modeling this in the Table model.

The effects of stylistic variation on the cognitive processing of a text has been a topic of debate within linguistics. We extend our understanding of how stylistic variation plays a role in modifying a reader's perception of a text into a new domain by looking at tense variation within restaurant reviews. Motivation for this stylistic variation is speculated to be driven by persuasive ability (i.e. to increase persuasive quality of a text). Several criteria have been identified to elicit narrative persuasion (emotionality, identification with the author, and transportation into a text), although there has not been a formal study to examine these factors in relation to the historical present. In this article we present two parallel studies designed to test how review sentences in past tense and historical present tense are perceived based on these criteria. We tested over 75 native English speakers of English asking them to judge either historical present or past tense negative restaurant review sentences based on four different criteria (persuasion, emotionality, identification, and transportation) using two different acceptability rating tasks (five-point scale). We find evidence of a relationship between the historical present and an increase in transportation and emotionality using a variety of statistical analysis techniques, including t-tests and regression analysis. We also find a relationship between past tense and identification. We interpret increased levels of emotionality and transportation as an indication that the historical present has the ability to influence a reader's cognitive processing of a text.

The role of contextual predictability in language processing and production has gained increasing attention in recent years, but in production remains relatively understudied above the phonetic/phonological levels. Further, the existence and source of predictability effects is still contested, and it remains unclear which phenomena these effects extend to, as well as what their explanatory range is (Jurafsky, 2003). I present two experiments which investigate the role of contextual predictability in the choice to omit optional subjects in Russian. The results demonstrate that when other predictors of referential expression choice are taken into account, contextual predictability, conditioned on preceding context, remains a significant predictor of whether an optional subject is pronounced or elided in colloquial text or speech. Contextual predictability conditioned on following context, in contrast, does not appear to exert significant influence after effects of preceding context are controlled for.

This thesis develops a compositional account of the contribution of steeply, monotonically rising and steeply, monotonically falling intonational tunes to the utterances they accompany. I pursue an account couched within a restrictive implementation of the Table model in which falling intonation signals that the speaker is making a commitment by virtue of their utterance, and rising intonation signals that the speaker is making no commitments by virtue of their utterance. I propose a bifurcation of the Table model into a doxastic portion of the context and a teleological portion of the context, and propose that imperatives interact with the teleological portion of the context, whereas declaratives and interrogatives interact with the doxastic portion of the context.

This dissertation seeks to address the following question: can we use prosodic meaning as a metric to inform the meaning of other semantically underspecified elements? This question is asked here with respect to a class of elements called discourse particles, which speakers use to structure a conversation in an expectationally and emotionally informative way. In this work, I propose to capture the effects of both prosody and intonation in an updated view of the Table Model of discourse representation (Gunlogson 2001, 2008, Farkas & Bruce 2010), wherein pragmatic effects are entered into the conversational scorecard by way of a speaker's Discourse Commitments. This view is underlain by a causal model framework of discourse structure (Halpern & Pearl 2000), which is based off of a probabilistic notion that I define as Expectation. Once these tools have been established, I show how this framework can be used to talk in an informed way about Mirativity (DeLancey 1997, Aikenvahld 2012). In doing this, I identify three discourse particles and one intonational contour in English that make up a part of the mirative system of this language. Broadening the reach, this work also picks out two strategies of discourse navigation in German whose effects can also be explained by making reference to speaker expectations in a discourse: the modal particles ja and doch. Here, the latter constitutes an example of a mirative marker in German, whose discourse effect picks out violated expectations of the speaker, while the former is an example of a particle that signals the status quo. This dissertation also seeks to experimentally validate the claims made about the interactions between prosody and discourse particle interactions. Is introduces a series of experiments designed to test the predictions that this new view on discourse management and coherence makes. In all, I argue that what these elements have in common is their ability to enrich a discourse with pragmatic material that contributes speaker-oriented commentary on a participant's beliefs and expectations in a conversation.

Learning a morphological process requires generalizing which words undergo the process and what form these words take. Emphatic reduplication in Turkish (TER) is unproductively applied to a small set of adjectives. Despite its unproductiveness, Turkish speakers have been shown to learn the form that adjectives that undergo emphatic reduplication take (Demircan, 1987; Wedel, 1999). This project investigates whether Turkish speakers learn which adjectives can undergo reduplication based on semantic regularities across TER adjectives. The experiments presented here will show that native Turkish speakers generalize over the color semantic class when learning the process of reduplication.

The purpose of this paper is to propose a new interpretation of doch, a German modal particle that has been widely studied, but whose contributions to an utterance are dis- puted by many prominent scholars. In what follows, I diverge from analyses put forth by Thurmair (1989), Abraham (1991), Rinas (2007), Grosz (2011, 2014), Rojas-Esponda (2013), and Müller (2014), and propose that doch’s contribution to an utterance rep- resents expectation violation with respect to a salient proposition in discourse. In par- ticular, the analysis here stems from two observations made about doch that is absent in previous work on this particle. First, I put forth evidence showing that doch is not only able to appear discourse-initially, a previously undiscussed property of this parti- cle, but also that the prejacent of doch need not be the propositional argument of the particle. Second, I argue that doch is a grammatical marker of mirativity, a category which inherently references surprise or violated expectation of a discourse participant (DeLancey, 1997, Aihkenvald, 2004). By taking these two steps, this approach is able to capture the previously undiscussed properties of this particle, and also explain an interesting distributional generalization about doch in polar questions.

Language comprehension requires a complex series of decisions under uncertainty. This is especially obvious when one string may have multiple different interpretations, whether due to lexical ambiguity, or the potential for an inference beyond literal content. This dissertation profiles how the human system for language comprehension times those decisions, specifically when and why it sometimes postpones them. Evidence comes from nine reading experiments in English probing variation across a range of different types of uncertain meaning (homonymy and polysemy, predicate distributivity, scalar implicatures from "some", and causal inferences from discourse coherence) and across two tasks (self-paced-reading and the Maze task of Forster et al., 2009). Diagnosing the presence of decisions by testing for garden-path effects and costs associated with selection against a bias, I highlight two key patterns. First, some decisions, e.g. sense specification for polysemous nouns, are delayed in normal reading, but occur immediately when a rapid decision would be more useful; I conclude that decisions to postpone are flexible and sensitive to a comprehender's goals. Second, possible scalar implicatures and causal inferences rapidly affect comprehension, but do not receive any typical, decisive commitment until much later; I conclude that comprehenders may develop expectations gradiently based on multiple possible interpretations before they make a firm decision. Throughout the dissertation, I explore how these and related facts might be explained as a consequence of the ways humans attempt to rationally allocate cognitive resources under uncertainty.

This dissertation explores, through three case studies, the relationship between the lexical semantics of clausal embedding (CE) predicates, and the syntactic and pragmatic restrictions on their complements, such as the (in)ability of certain predicates to embed declarative clauses or nominal expressions. I propose that these restrictions arise from interactions between fine-grained aspects of predicate meaning and the linguistic environment in which those predicates occur, rather than being stipulated into predicates' lexical entries via restrictions on permissible semantic types or semantic categories of their arguments (e.g. Grimshaw 1979, Pesetsky 1982, 1991) or resulting from polysemy or ambiguity of CE predicates (e.g. George 2011). First, I examine the Estonian verb mõtlema, which has a believe-like interpretation with an embedded declarative, and a wonder-like interpretation with embedded declaratives. I show that this behavior can be derived straightforwardly only if declaratives and interrogatives are typewise identical, i.e. sets of sets of worlds (Hamblin 1973, Ciardelli et al. 2013), and \textit{mõtlema} denotes an ontologically primitive attitude of 'contemplation'. Second, I analyze the behavior of English believe, which cannot embed interrogatives except under a combination of modal and nonveridical operators, such as can't. I propose that the apparent 'lifting' of selectional restrictions of believe in some contexts, as well as other unexpected properties of the can't believe construction, can be understood compositionally: believe can in fact compose with interrogative clauses in principle, but doing so normally results in systematically trivial meanings (Theiler et al. 2019). By placing believe under the right combination of operators, this triviality can be obviated (see Mayr 2019). Finally, I develop a semantic account of verbs like believe and trust, which can embed variety of nominal expressions alongside clauses, posing a compositional puzzle (Djärv 2019). I propose that these predicates are a kind of weak response-stance predicates (Cattell 1978), in that they presuppose an evidential source of the relevant attitude, which can be spelled out as a direct object. I take this to suggest a tight link between argument structure of belief predicates and their external syntactic distribution.

Responsive Predicates (ResPs)–those clausal-embedding predicates which permit both declarative or interrogative complements, like know and say–pose a puzzle for compositional semantics. If declarative clauses denote propositions and interrogative clauses denote questions, it is difficult to explain how ResPs may select both types of clauses. Prior accounts diverge on how to best rectify this inconsistency, from reducing interrogatives to propositions (Karttunen 1977, Ginsburg 1995, Spector & Egré 2015, a.o.), to reducing declaratives to questions (Uegaki 2016) or dispensing altogether with the assumption that the denotations of declaratives and interrogatives are of different types, as in Inquisitive Semantics (e.g. Theiler et al 2016).

In this thesis I bring novel data from Estonian to bear on this theoretical debate. In Estonian, a class of ResPs whose denotations concern contemplation, such as mõtlema ’think, consider’, convey what radically different meanings depending on the type of

complement it takes: When paired with a declarative complement, mõtlema canonically indicates representational belief like English think, but with an interrogative complement, it indicates ignorance toward the true answer to the embedded question, like English wonder.

I demonstrate that a proposition-embedding denotation for mõtlema is not compatible with its full range of meanings. Instead, I propose a question-embedding semantics for motlema, which relates an individual to the set of issues they are contemplating. I show how this denotation, combined with standard Gricean reasoning, can naturally derive the superficially distinct interpretations of mõtlema across contexts. Building upon the analysis by Uegaki (2016), I argue that a proposition-embedding semantics for ResPs is not sufficient to capture the Estonian facts, whereas a question-embedding semantics is. Furthermore, I suggest that this approach is also compatible with the uniform treatment of declaratives and interrogatives of Inquisitive Semantics.