The neural mechanisms of predictive coding during language comprehension

在熟练的阅读过程中，人们从书面字符串中识别单词，并将单词联合收割机组合成更大的消息级解释。阅读是现代人类生活中普遍存在的一部分，受损的阅读对学业和职业成功非常不利。尽管它的重要性，使熟练的读者理解多词组合的神经过程仍然知之甚少。在这项研究中，科罗拉多大学博尔德分校的阿尔伯特·金博士和菲利普·吉利博士将使用尖端的神经生理学方法来了解大脑如何在阅读过程中产生和使用关于文本中即将出现的单词的预测，并由先前的语言背景指导。预测被认为支持正常阅读的快节奏，通常是每秒3-5个单词，通过允许大脑在预期的文本到达视网膜输入之前提前准备。预测也被认为是至关重要的共同能力，准确地感知单词在贫困的输入，如条件下的照明，打印质量，或视觉敏锐度。该项目的研究预计将产生指导识别阅读障碍和区分不同类型障碍的知识，从而指导阅读障碍的临床和教学方法。该项目将使用头皮记录的脑电描记术（EEG）观察神经振荡活动-反映神经网络的动态耦合和解耦-而健康的年轻人阅读句子。根据计算语言模型，参与者阅读的句子将被操纵，以便某些句子上下文使特定的单词可能出现。 在这些条件下，将检查在语言输入中出现单词之前和之后发生的大脑活动，以寻找预测的证据。该项目将研究神经振荡在将关于即将到来的单词的预测从高级别大脑区域传达到低级别视觉皮层区域以及评估预测与自下而上的感觉输入之间的匹配方面的作用。研究人员还将研究参与这种预测活动的语言背景和处理需求的类型。这项研究的结果预计将有助于显着的认知神经科学相结合的三个关键领域的研究，通常已经分别研究：语言处理，预测认知和神经振荡。该项目将开发新的分析方法，用于表征EEG数据中的神经振荡，这些方法将被开发成一个工具箱，与其他研究人员共享。该项目还将产生一个丰富的数据库，在阅读自然故事的过程中的大脑活动，这将是与科学界共享作为进一步研究的资源。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

During skilled reading, people recognize words from written character strings and combine words into a larger message-level interpretation. Reading is a pervasive part of modern human life, and impaired reading is highly detrimental to scholastic and professional success. Despite its importance, the neural processes that enable skilled readers to comprehend multi-word combinations remain poorly understood. In this research, Dr. Albert Kim and Dr. Phillip Gilley of the University of Colorado Boulder will use cutting edge neurophysiological methods to understand how the brain generates and uses predictions about the upcoming words within a text during reading, guided by prior linguistic context. Predictions are thought to support the fast pace of normal reading, which is typically 3-5 words per second, by allowing the brain to prepare in advance for anticipated text before it arrives in the retinal input. Predictions are also thought to be crucial to the common ability to perceive words accurately in the face of impoverished inputs, such as conditions of poor lighting, print quality, or visual acuity. The research in this project is expected to produce knowledge that guides the identification of reading disorders and distinctions between different sorts of disorders, which can in turn guide clinical and pedagogical approaches to reading disorders.The project will use scalp-recorded encephalography (EEG) to observe neural oscillatory activity -- reflecting the dynamic coupling and uncoupling of neural networks -- while healthy young adults read sentences. The sentences that participants read will be manipulated so that some sentence contexts render a specific word likely to occur, according to computational language models. Under these conditions, brain activity that occurs before and after the appearance of a word in the linguistic input will be examined for evidence of predictions. The project will investigate the role of neural oscillations in conveying predictions about upcoming words from high level brain regions to low level visual cortical areas and also in assessing the match between predictions and the bottom-up sensory input. The researchers will also study the types of linguistic contexts and processing demands that engage such predictive activity. Results of this research are expected to contribute significantly to cognitive neuroscience by combining three critical areas of research that have typically been studied separately: language processing, prediction in cognition, and neural oscillations. The project will develop new analytic methods for characterizing neural oscillations in EEG data, which will be developed into a toolbox to be shared with other researchers. The project will also produce a rich database of brain activity during the reading of naturalistic stories, which will be shared with the scientific community as a resource for further research.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.