This chapter outlines a framework for using signal detection theory (SDT) to guide the design and analysis of acceptability judgment studies in experimental linguistics. It presents a worked example experiment on the syntactic phenomenon of D-linking (discourse linking) and wh-movement. It shows how to derive common SDT measures (like d_sub_a and s), how to do inferential statistics over those measures, and how to find additional theoretical and practical resources.

This thesis investigates syntactic and thematic factors that contribute to interference in language comprehension during the resolution of subject-verb dependencies. One self-paced reading experiment and one speeded acceptability judgment task examined the mechanisms underlying subject retrieval operations using materials that leverage syntactically alternating Double Object (DO) and Prepositional Dative (PP) structures. The self-paced reading results suggest that arguments in the PP frame inhibit retrieval during thematic binding operations, but identical arguments in the DO frame do not. The acceptability judgment results indicate that arguments in both structures can engender facilitatory interference during number agreement processing in ungrammatical sentences. This thesis proposes an account of the present results, as well as a range of previous findings, involving a single set of retrieval cues, weighted with estimates of reliability according to the grammatical function of the retrieval operation. Such an account is consistent with a cue-integration model of language comprehension (Martin 2016), which links psycholinguistic theory with neurobiological models of perception. Additionally, this work demonstrates both inhibitory and facilitatory interference effects with identical structures, providing further evidence for central predictions of the cue-based retrieval model of sentence processing (Lewis & Vasishth 2005).

The main question proposed in this study concentrates on the issue of whether implicit agents trigger retrieval interference in sentence comprehension and whether this depends on their syntactic representation. In order to address this question, we examined two impersonal passive constructions in the Ukrainian language, which differ with regard to the presence of implicit causative agent. This study presents an online forced-choice acceptability and self-paced reading experiments comparing the processing of the passive constructions with syntactically represented implicit agents and passive constructions without such implicit agents. In addition, the online self-paced reading experiment compares the processing of subject extracted relative clauses (SRCs) and object extracted relative clauses (ORCs) to reveal whether parsing relative clause effects retrieval interference and test whether a widely known ORC/SRC asymmetry shown in many languages including Russian (Levy, 2013) holds for Ukrainian as well. The results of the self-paced reading experiment support a retrieval-based account of interference effects in sentence processing and the idea that sentence processing is constrained by working memory limitations. Conducting this study contributes to the existing psycholinguistic literature on retrieval interference and processing of implicit agents, and helps us gain a better understanding of the nature of the parsing mechanisms that support human language processing.

English resumptive pronouns, as in "...the flowers that I don't know where IT came from," are enigmatic in that they are judged to be unacceptable, which would indicate that they are ungrammatical, but are regularly produced by native speakers, which is typically taken to indicate grammaticality. We report results from two studies: an acceptability judgment study on sentences with resumptive pronouns or gaps ("...the flowers that I don't know where _came from"), and a written production study which elicited sentences that required participants to produce either a gap or a resumptive pronoun in various island and non-island domains. We find that, in a given structure, resumptive pronouns are produced at a rate that negatively correlates with the acceptability of the corresponding structure with a gap in it. That is, where gaps are less acceptable, resumptive pronouns are more common. To account for these data, we offer a model of English production processes as sensitive to the acceptability of a planned utterance. When the system detects impending unacceptability, it may give up on the global plan to form a syntactic dependency. When this happens, a gap is no longer licensed and a pronoun is used to satisfy local subcategorization constraints.

This work focuses on the processing of wh-in-situ dependencies in Turkish to understand the mechanisms of the search process for the scope position of a wh-in-situ element. Despite the lack of overt movement, it is argued that there is a covert relationship between the SPEC CP position of the scope-bearing clause and the wh-element. Previous work investigating the processing of wh-in-situ elements finds evidence for an abstract dependency between the scope position and the wh-element. This study focuses on the search process for upcoming structural information, specifically, whether the clause most local to a wh-element is expected to bear scope. Since Turkish only marks the scope of the wh-element via intonation, a reading study on wh-in-situ can allow us to directly probe the relationship between the two ends of the dependency. A self-paced reading study shows that embedding clauses that do not allow local scope incur a slowdown in reading times when they contain a wh-element. A follow-up fragment completion study shows an overwhelming preference for local scope when participants are provided with fragments containing a wh-element. The results are interpreted as a locality preference during the production and processing of wh-in-situ dependencies.

A long-standing question concerning ellipsis considers the nature of the representations that mediate interpretation at ellipsis sites. The present study asks what kind of material contained in the representation is used in real-time ellipsis processing.

A probe is needed to investigate the nature of the reactivated representation. Number features are used, as these can feed agreement attraction (AA), a process sensitive to morpho-syntactic information. In AA, an intervening, plural NP—the attractor—agrees with the verb as opposed to the singular, head noun containing it.

[The key(Head Noun) to [the cabinets(Attractor)]PL]SG werePL on the table.

When Noun Phrase Ellipsis (NPE) elides an AA-triggering NP, would an attractor in the antecedent generate attraction when reactivated at the ellipsis site? Depending on how exhaustive reactivation is, different sets of number features may be reactivated. If the antecedent were reactivated fully—including the attractor—attraction is predicted; but if reactivation is only partial—just the head noun—no attraction is predicted. Based on the results of four reading-time experiments examining NPE, the data support that reactivation at ellipsis sites is syntactic and partial.