Doctoral Dissertation Research: Resolution in context: Identity and nonidentity in anaphora resolution

本论文的研究目的是通过研究两种语言结构与周围语言和非语言语境的关系，提高对句子产生和理解的科学认识。语言学家长期以来一直意识到，说话者对在话语的广泛语境中可获得的信息的意识会影响他们如何使用和解释语言，但这种知识对语言系统某些部分的确切影响尚未完全确定。本文研究的结构是动词短语省略，即助动词后的无声“间隙”被解释为包含完整的言语内容；降音，即句子成分的发音被淡化，表明它是多余的。先前的分析表明，这两种现象只有在它们可以被解释为与语言话语中较早出现的显性先行词相同时才是合法的。这篇论文调查了一些证据，这些证据表明，以前对这些现象的同一性表述过于严格，它们可以在更广泛的语言先行词以及非语言语境中可用的信息方面得到许可和解释。除了促进对这些现象的科学理解外，这项研究可能具有技术和临床应用价值。更全面地了解听者如何整合语言和非语言信息来解释这些结构，可能会导致人工语言系统的发展，从而产生和解释与人类说话者更相似的语言，以及对使用这些结构受损的说话者的诊断和治疗策略。论文项目主要围绕三个主要研究问题：(1)当存在可以提供合理解释的语言先行词时，在动词短语省略的解释中是否考虑了非语言语境中的突出信息？(2)除重复以外的语言关系（即蕴涵、桥接或世界知识推理）是否允许特定成分的降音？(3)通过在非语言语境中突出的信息，可以全部或部分地允许降音吗？这两种现象将使用生产、感知和眼动追踪方法进行调查，这些方法将检查动词短语省略和降音的使用和解释，并在离线和实时进行。通过考虑语言诱发信息、可推断信息和突出的非语言信息，研究结果将增强这些结构如何在更全面的话语语境中被许可和解释的模型，而不是之前所考虑的。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The goal of this dissertation research is to enhance the scientific understanding of sentence production and comprehension by examining the relationship between two types of linguistic construction and the surrounding linguistic and nonlinguistic context. Linguists have long been aware that speakers' awareness of what information is available in the broad context of utterance affects how they use and interpret language, but the exact effects of this knowledge on certain parts of the linguistic system have not been fully identified. The constructions under investigation in this dissertation are verb phrase ellipsis, in which a silent "gap" after an auxiliary verb is interpreted as containing full verbal content, and deaccenting, in which the pronunciation of a sentential constituent is de-emphasized to indicate that it is redundant. Previous analyses have suggested that both phenomena are licit only when they can be construed as identical to an overt antecedent that appears earlier in the linguistic discourse. This dissertation investigates evidence that previous formulations of identity for these phenomena have been too strict, and that they can be licensed and interpreted with respect to a broader range of linguistic antecedents, as well as with respect to information available in the nonlinguistic context. In addition to advancing the scientific understanding of these phenomena, this research may have technological and clinical applications. A more complete understanding of how listeners integrate linguistic and nonlinguistic information to interpret these constructions may lead to the development of artificial language systems that produce and interpret language more similarly to human speakers, as well as diagnosis and treatment strategies for speakers whose use of these constructions is impaired. The dissertation project focuses on three main research questions: (1) Is salient information from the nonlinguistic context considered during the interpretation of verb phrase ellipsis when there is a linguistic antecedent that could provide a reasonable interpretation? (2) Can deaccenting of a particular constituent be licensed by linguistic relations other than repetition, i.e., entailment, bridging, or world knowledge inferences? (3) Can deaccenting be licensed, in whole or in part, by information made salient in the nonlinguistic context? The two phenomena will be investigated using production, perception, and eye-tracking methodologies that examine the use and interpretation of verb phrase ellipsis and deaccenting both offline and in real time. The results will enhance models of how these constructions are licensed and interpreted in a more comprehensive discourse context than has previously been considered, by taking into account linguistically evoked information, inferrable information, and salient nonlinguistic information.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.