CAREER: Integrating Grammatical and Psycholinguistic Approaches to Phonological Processes in Speech Production

该奖项由2009年美国复苏和再投资法案(公法111-5)资助。音系学研究构建了丰富的数学理论，准确地表征了人类语言中观察到的声音的系统模式。然而，这些理论在很大程度上忽略了人类如何实时处理声音。例如，当说话者决定发出特定的声音序列时，他如何计划嘴唇和舌头的运动序列以产生适当的声音？这些问题一直是心理语言学研究的焦点，在那里，详细的实验导致了关于人类声音处理的复杂但基本上是定性的理论。该项目整合了正式语言学和心理语言学方法的长处，为人类言语产生的研究开发了一个更强大的框架。梯度调和语法理论是建立在最近发展起来的调和语法的基础上的，调和语法是一种数值化的、基于约束的声音结构理论。该项目将通过纳入梯度--表征和加工机制中的非范畴变量--来扩展这一理论，这是心理语言学的一个核心概念。梯度调和语法将通过一组量化模型进行测试，这些模型检查来自仔细、无错误的语音和包含发音错误的语音的实验数据。然后，框架和计算工具将被扩展以模拟双语说话者中多个声音系统的梯度相互作用。为了让研究界最大限度地利用梯度调和语法框架，将在因特网上公开提供一套灵活的、用户友好的、开放源码的应用程序。这些应用程序将包括一组实验室作业，以帮助教育工作者将这些工具带入课堂。关于言语错误的实验工作可以作为未来对言语产生障碍的研究的基础，例如中风患者的言语产生障碍。其他研究也将使用这个一般框架来帮助理解双语说话者在演讲过程中是如何管理多个声音系统的。在一个日益多语言、相互联系的世界里，这样的研究是理解人类语言的关键部分。

This award is funded under the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).Research in phonology has constructed rich mathematical theories to precisely characterize the systematic patterns of sounds observed in human languages. However, such theories have largely ignored how humans process sounds in real time. For example, when a speaker decides to produce a particular sequence of sounds, how does he plan out the sequence of movements of his lips and tongue to produce the appropriate sounds? These questions have been the focus of research in psycholinguistics, where detailed experimentation has led to sophisticated but largely qualitative theories of human sound processing. This project integrates the strengths of formal linguistic and psycholinguistic approaches to develop a more robust framework for the study of human speech production. The theory, Gradient Harmonic Grammar, builds on the recently-developed Harmonic Grammar, which is a numerical, constraint-based theory of sound structure. The project will extend this theory by incorporating gradience--non-categorical variation in representational and processing mechanisms--which is a central concept in psycholinguistics. Gradient Harmonic Grammar will be tested via a set of quantitative models that examine experimental data from careful, error-free speech and speech containing mispronunciations. The framework and computational tools will then be extended to model the gradient interaction of multiple sound systems in bilingual speakers. To make the Gradient Harmonic Grammar framework maximally accessible to the research community, a set of flexible, user-friendly, open-source applications will be made publicly available over the internet. These applications will include a set of laboratory assignments to help educators bring these tools into the classroom. The experimental work on speech errors can serve as the basis for future research into speech production disorders, for example, those of stroke victims. Additional studies will also use this general framework to help understand how bilingual speakers manage multiple sound systems during speech proudction. Such research is a critical part of understanding human language in an increasingly multilingual, interconnected world.