The human processing of natural language occasionally reflects systematicies of the system’s formal organization, often at the level of syntax. Data on the processing of appositive relative clauses (Dillon et al., 2014; Dillon et al., 2017) suggest that the processor may also be sensitive to the pragmatic organization of information within a sentence: independent speech acts embedded within a maximal speech act appear to be isolated from later parsing and decision making. In three experiments, I examine the processing of another structure which may create an embedded speech act, direct discourse speech reports. Despite existing evidence suggesting direct discourse is indeed processed independently from its superordinate context, the experiments reported here find no support for the prediction that it is isolated from downstream processes. Taking this as evidence against the strong hypothesis that subordinate speech acts serve as hard boundaries in processing, I outline possible alternative hypotheses.

This dissertation is about implicit prosody, the prosodic structure that readers assign during silent reading. The dissertation has several goals: determining which reading tasks are appropriate for studying implicit prosody, establishing how grammatical principles could guide incremental assignment of prosodic structure, and investigating how implicit prosody interacts with other properties such as focus in order to influence syntactic parsing and interpretation. On the methodological side, this dissertation demonstrates that the Maze task is suitable for studying implicit prosody by replicating several major findings on metrical and phrasal prosody in both the Maze task and in self-paced reading. In addition to showing that the Maze is sensitive to implicit prosody, the methodological comparison confirms previously reported advantages of the Maze over self-paced reading, such as more localized and larger effects. On the theoretical side, the dissertation lays the groundwork for developing an incremental model of prosodic parsing. I provide an overview of the major grammatical constraints that govern the syntax-prosody interface, drawing on work from the theoretical phonology literature. I discuss how and when these grammatical constraints, which are typically invoked to model the final phrasing for a complete sentence structure, could be deployed by an incremental parser that assigns a prosodic structure word-by-word. Using a toy model of an incremental parser, I also show how the parser’s first pass implicit prosody may differ from the final prosody, arguing that future work in this area should more closely consider these potential differences. The final set of experiments investigates both the timing of implicit prosodic assignment and how prosodic structure and information structure affect attachment decisions. Based on the results of these experiments, I propose the Visibility First Hypothesis, according to which attachment decisions are determined primarily by prosodic visibility, while other factors such as focus only exert an influence when two potential attachment sites are equally visible. I then outline several experiments to test the Visibility First Hypothesis in future work.

The goal of this thesis is to better understand the impact of different phonological feature classes (voice, place of articulation, manner of articulation) on the likelihood of a listener recognizing a mispronounced word in natural speech. I examine this question by introducing mispronunciations into binomial expressions, semi-idiomatic phrases like salt and pepper, where the first half of the phrase lexically and semantically primes the second half. Mispronunciations were produced by deliberately changing the voice, place, or manner features of the onset consonant of the third word in a binomial. A set of experiments investigates (i) how listeners rate the effect of different feature errors on overall pronunciation quality, and (ii) how accurate they are at correctly recalling those erroneous pronunciations. Results are first analyzed for an effect of mispronunciation and find that listeners give higher ratings and are more accurate at correctly recalling binomial expressions with no mispronunciation, regardless of the particular feature change. A subsequent set of post-hoc analyses of the results are run, comparing the different mispronunciation conditions against each other and breaking down and comparing the distinctive features that comprise the manner feature class (continuancy, stridency, nasality, lateralization). Results of the ratings and recall task find that different feature changes differently impact listener perception of sound errors: listeners are less likely to make mistakes when recalling words with manner errors in the target position than words with voice or place errors, and they assign lower ratings to binomial expressions with manner feature errors than equivalent binomials with mispronunciations involving voicing features.

This dissertation aims to show how existing behavioral evidence regarding the processing of focus can be brought more in line with an alternative-based understanding of focus as proposed in theoretical semantics, without loosing sight of the way general comprehension pressures may shape its interpretation. Throughout, I argue that this is possible by studying how the mental representations involved in the processing of focus are incrementally constructed during sentence comprehension. Reading measures obtained in a series of Maze experiments show that comprehenders arrive at a final interpretation of a focus by combining multiple sources of evidence, including lexical, conceptual and world-knowledge, as well as fine-grained linguistic representations that guide the incremental interpretation of focus independently from such general knowledge. These findings allow for a unified understanding of the inconsistent results previously found in the reading of focus, while also explaining how alternatives and discourse context are involved in the prioritization and anticipation of foci.

Experiments 1-4 show, first, that the comprehension of focus generally induces a processing cost because reading times on foci are longer than on non-foci. It then shows that this cost is reduced when contrastive alternatives to the focus are mentioned in the preceding context, suggesting that the representation of contrastive alternatives is indeed involved in the comprehension of focus. The presence of focus marking induces a cost that is separable from a cost of interpreting newly introduced information, and that the presence of alternatives provides a reading benefit that is separable from a benefit due to semantic priming. Together, these findings suggest that contrastive alternatives must somehow be involved in the processing of focus, and that its cost cannot be explained in terms of a general cost for new material.

Experiments 5-7 investigate how discourse context is used to assign a focus structure to a sentence in incremental sentence comprehension. It shows that the presence of contrasting material in the context is used by comprehenders to assign focus marking to subsequent sentences, suggesting that representations of contextual contrasts are utilized to anticipate the location of an upcoming focus. Again, results indicate that these behavioral effects of focus are separable from effects of newness or the predictability of upcoming material in general. This suggests that it is crucially the fact that comprehenders encode abstract representations of contextual contrasts that gives rise to these behavioral effects of focus, not the presumed communicative importance of foci or unpredictability alone.

Finally, experiments 8-10 study what information comprehenders rely on in constructing an alternative set to a focus. This chapter again provides evidence for the claim that abstract linguistic representations of the discourse context are used to either rule in or rule out potential members of the alternative set to a focus. It shows that the deployment of these types of representations is fast, and independent from the use of general conceptual knowledge or the use of domain-general mechanisms such as semantic priming. I thus propose that comprehenders rapidly revisit semantic representations of the discourse context in constructing the alternative set to a focus.

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 is about how discourse is comprehended in real-time, and how it is represented in memory. This project begins to bridge the gap between formal linguistic theories of discourse coherence and structure and psychological theories of discourse processing. I argue that combining the insights of these two strands can bring us to a deeper understanding of the core puzzles in discourse research, e.g., what determines whether or not a sequence of clauses or sentences seems like a coherent discourse? What factors help or hinder the on-line comprehension, the incremental interpretation, of discourse? How is discourse structured in memory, and to what degree does this structure resemble formal linguistic representations of discourse? I approach these questions via three sets of experiments. First, I investigate the role of temporal information in on-line discourse comprehension, give special scrutiny to the common assumption that progression is the default temporal relation in discourse processing. My results suggest that, if progression is any sort of default, it is not the sort that has a strong influence on on-line processing. The second set of experiments is situated in the context of Situation Model Theory, a prominent psychological theory of discourse. I investigate the roles of temporal and causal information in the construction of discourse structure in memory. Based on my findings, I argue that the influence of temporal factors may not be as clear-cut as previous work has suggested. I also argue that causal factors have a relatively large role in this domain. Finally, I turn to Segmented Discourse Representation Theory, a prominent linguistic coherence theory, and its Right Frontier Constraint, which is argued to be the primary governor of pronominal anaphora in discourse. I challenge this view, arguing that the effects that are attributed to this constraint can in fact be accounted for by independent factors that influence the salience of entities in discourse.

Studies on the reading of appositive relative clauses (ARCs) have found that ARCs seem to exhibit less influence in later parsing and decision-making than similar constructions (Dillon et al. 2014, 2017), a pattern we call discounting. Existing accounts often link discounting to the status of ARCs as independent segments in systems of pragmatic representation. This would predict discounting for other constructions as well. In this study, we test that prediction by investigating the reading of direct discourse speech reports and causal adjuncts in English. Diagnostics supplied by the theoretical literature show that these constructions contribute the same independent segments as ARCs in two different systems of pragmatic organization: direct discourse reports contribute an independent speech act, and causal adjuncts contribute their own discourse units. Nevertheless, in a series of five experiments, we find no evidence of ARC-like discounting for either. We conclude that discounting should not be linked to either of these pragmatic representations, and discuss the outlook for other approaches to the phenomenon.