Computational and Neurophysiological Dynamics of Lexical Access in Speech Perception

项目概要/摘要 从声音到意义的映射，是理解口语的基础 涉及一系列包括两个基本认知子程序的神经计算：感知处理， 其涉及语音信号的低水平听觉和语音分析以及词汇访问， 检索与词的声音形式相关联的语义和其它属性。虽然过去的工作 研究了知觉加工和词汇通达的神经解剖学基础， 这些限制阻止了研究人员探索实现这些功能的神经生理学动力学。 在细粒度的空间和时间尺度上进行计算。获取和分析的最新进展 直接来自清醒、行为正常的人类大脑的高密度颅内皮质电图（ECoG）数据 接受脑外科手术的科学家们已经阐明了语音的皮层表征和 知觉处理，但目前还不知道大脑如何实现词汇访问。在拟议的工作中， 我们将探讨词汇通达的神经生理学以及声音和意义之间的相互作用， 记录ECoG对口语的反应，同时植入电极网格的患者参与两个 心理语言学实验（目标1），并通过开发一个计算模型，连接内部模型 预测时空分布的处理组件的神经响应语音（目标2）。 目的1.1和1.2将检验不同的情境偏差对神经生理反应的影响 一个口头的话。在目标1.1中，我们假设声学线索和语境线索都能调节词汇 词汇通达的神经信号将出现在上级颞回（STG）。 重要的是，在不同的环境中，对声学上相同的刺激的ECoG反应的任何一致差异， 上下文将对应于与词汇访问（而不是感知）相关联的神经处理组件 加工）。在目标1.2中，我们假设自上而下的语境偏见会调节知觉加工 随后的相关口语单词，符合高度互动的感知和词汇处理 STG中的动态最后，在目标2中，我们将开发一个能够 预测时变神经反应模式这样一个模型的规范将正式形成一个理论 框架，在其中考虑有关感知处理的神经实现的假设， 词汇存取它还将产生新颖的，可测试的预测，将指导未来的实验工作， ECoG涉及言语感知。了解支持的神经电路和计算 知觉加工和词汇通达有可能改善诊断和表征 言语和语言障碍（例如，失读症，失语症），导致创新的，基于大脑的治疗方法， 条件，并帮助开发能够从语音中提取含义的辅助技术。

Project Summary/Abstract The mapping from sound to meaning that is fundamental to the comprehension of spoken language involves a series of neural computations comprising two basic cognitive subroutines: perceptual processing, which involves the low-level auditory and phonetic analysis of a speech signal, and lexical access, whereby the semantic and other properties associated with a word's sound form are retrieved. While past work has investigated the neuroanatomical bases for perceptual processing and lexical access, methodological limitations have prevented researchers from probing the neurophysiological dynamics that implement these computations on a fine-grained spatial and temporal scale. Recent advances in the acquisition and analysis of high-density intracranial electrocorticography (ECoG) data directly from the brains of awake, behaving humans undergoing brain surgery have elucidated the cortical representations of speech sounds and the dynamics of perceptual processing, but it is not yet known how the brain implements lexical access. In the proposed work, we will investigate the neurophysiology of lexical access and the interactions between sound and meaning by recording ECoG responses to spoken words while patients with implanted electrode grids participate in two psycholinguistic experiments (Aim 1), and by developing a computational model that links internal model predictions to spatiotemporally distributed processing components in neural responses to speech (Aim 2). Aims 1.1 and 1.2 will examine the influence of different contextual biases on the neurophysiological response to a spoken word. In Aim 1.1, we hypothesize that both acoustic and contextual cues will modulate lexical access, and that the neural signature of lexical access will emerge in the superior temporal gyrus (STG). Importantly, any consistent differences in ECoG responses to an acoustically identical stimulus in different contexts will correspond to neural processing components associated with lexical access (and not perceptual processing). In Aim 1.2, we hypothesize that top-down contextual biases will modulate perceptual processing of a subsequent related spoken word, consistent with highly interactive perceptual and lexical processing dynamics in STG. Finally, in Aim 2, we will develop a computational model of speech perception capable of predicting time-varying neural response patterns. The specification of such a model will formalize a theoretical framework within which to consider hypotheses about the neural implementations of perceptual processing and lexical access. It will also generate novel, testable predictions that will guide future experimental work with ECoG involving speech perception. Understanding the neural circuitry and computations that s upport perceptual processing and lexical access has the potential to improve diagnosis and characterization of speech and language disorders (e.g., dyslexia, aphasia), lead to innovative, brain-based treatments for such conditions, and aid in the development of assistive technologies capable of extracting meaning from speech.