Representation, Storage, and Decay in Auditory-Verbal Working Memory

这项研究的总体目标是显著推进我们对言语工作记忆神经机制的理解。几十年来，对言语工作记忆的认知心理学研究表明，人们记住言语信息的主要方式是通过无声的排练，或内部言语。换句话说，为了让语言信息在意识中保持高位，我们利用自己的能力与自己进行内部对话。在Baddeley及其同事的经典言语工作记忆模型中，语音回路，内部言语被分解为两个通信组件：运动/发音过程和语音/听觉组件。在过去的20年里，认知神经科学试图了解这两个组成部分是如何“映射”到大脑的。然而，我们仍然对言语工作记忆的机制基础缺乏了解。因此，虽然我们知道在在线维护语言信息的过程中，运动和听觉皮层区域都是“活跃的”，但我们对这些区域进行的计算却知之甚少。 在这个提议中，我们使用一系列的功能性磁共振成像（fMRI）研究来阐明语言记忆是在大脑的听觉皮层编码。随着多体元模式分析（MVPA）技术的出现，现在可以捕获并可靠地识别编码特定刺激或认知状态的神经激活的分布式模式。在这里，我们利用功能磁共振成像技术的这一非凡发展来“解码”听觉皮层的活动模式，这些活动模式在听觉记忆提取和内部言语过程中表现出来。首先，我们将测试这一假设，即外部和内部言语诱发的活动模式在听觉皮层中共享一个表征基础。第二，我们将测试一个关于内部言语的著名解释，它表明内部言语是运动到感觉放电的结果，这是内部监测回路的一部分，对言语产生中的错误纠正很重要。最后，我们将探讨如何“表面特征”的感知语音，如性别和口音，被纳入听觉记忆痕迹，在工作记忆过程中被激活。 虽然这一建议的重点是在词汇-言语模态，其影响超出了这个“模型子系统”，因为我们从目前的一组实验中学到的东西将是相关的，并适用于其他记忆域。此外，这项工作的一个关键目标是继续追求一套统一的原则，这些原则是不同类型记忆（包括言语记忆、视觉空间记忆、短期记忆和长期记忆、情景记忆和语义记忆）的存储、表征和激活的基础。因此，这项工作是一个更大的跨职业计划的一部分，其长期目标是对人类记忆进行更连贯、更少分割的理解。

The overarching goal of the proposed research is to significantly advance our understanding of the neural mechanisms underlying verbal working memory. Decades of research on the cognitive psychology of verbal working memory have shown that the primary way in which people keep verbal information in mind is through subvocal rehearsal, or inner speech. In other words, to keep verbal information aloft in consciousness we make use of our capacity to carry on an internal conversation with ourselves. In Baddeley and colleagues' classic model of verbal working memory, the phonological loop, inner speech is decomposed into two communicating components: a motor/articulatory process and a phonological/auditory component. In the last 20 years, cognitive neuroscience has tried to understand how these two components "map" on to the brain. However, we still have a poor understanding of the mechanistic basis of verbal working memory. Thus, while we know that during the online maintenance of verbal information both motor/ and auditory cortical areas are "active", we have a poor of the computations carried out by these regions. In this proposal we use a series of functional magnetic resonance imaging (fMRI) studies to elucidate memories for speech are coded in the brain's auditory cortex. With the advent of multivoxel pattern analysis (MVPA) techniques, it is now possible to capture and reliably identify distributed patterns of neural activation that code for particular stimuli or cognitive states. Here we take advantage of this extraordinary development in fMRI technology to "decode" the patterns of activity in auditory cortex that are manifested during auditory memory retrieval and inner speech. First, we will test the hypothesis that patterns of activity evoked by external and inner speech share a representational basis in auditory cortex. Second, we will test a prominent account of inner speech that suggests that it emerges as a result of motor-to-sensory discharges that are part of an internal monitoring circuit important for error correction in speech production. Last, we will investigate how "surface features" of perceived speech, such as gender and accent, are incorporated into the auditory memory traces that are activated during working memory. Although the focus of this proposal is in the auditory-verbal modality, the implications go beyond this "model subsystem" because the what we learn from the current set of experiments will be relevant and applicable to other memory domains. Furthermore, a key aim of this work is to continue to pursue a set of unifying principles that underlie the storage, representation, and activation of different kinds of memory, including verbal, visuospatial, short- and long-term, episodic and semantic. This work, therefore, is part of a larger career-spanning program with the long-term goal of devising a more coherent and less compartmentalized understanding of human memory.