题 目：How Students Understand STEM Concepts: An fMRI study of Conceptual Representation through Expository Text Comprehension
李平，北京大学中文语言学学士 (1983)，普朗克研究所心理语言学及荷兰莱顿大学博士。毕业后在加州大学圣地亚哥分校的语言研究中心和McDonald Pew认知神经科学中心进行博士后研究。 1992年至1996年期间担任香港中文大学助理教授， 2004-2008年担任美国里士满大学心理学和认知科学教授。目前他是宾夕法尼亚州立大学心理学、语言学、信息科学与技术等专业终身教授，脑、行为、认知中心主任，计算科学研究院副主任。他还是美国白宫脑计划美国科学基金会科研课题首席科学家 (2015-2019)， 中国长江学者讲座教授 (2015-2017)。目前担任Journal of Neurolinguistics的主编，Frontiers in Psychology:Language Sciences的副主编。他曾担任Bilingualism: Language andCognition 的主编 (2003-2013)，Society for Computers in Psychology的主席(2012)及美国国家科学基金会的认知神经科学计划和感知、行为和认知计划的主任(2007-2009)。 李平在心理语言学、语言习得、计算机模型及双语的认知神经机制等领域出版了大量的研究专著及150余篇学术论文。有关李平教授研究的更多信息请点击http://blclab.org/。
In the past decade we have seen a surge of interest in using neuroimaging methods, particularly functional MRI, to study text and discourse comprehension. Exciting new opportunities have arisen for our further understanding in this domain due to methodological advances in using multimodal imaging, including technological development in combining neuroimaging with eye-movement methods and theoretical understanding into the interaction between individual differences in cognitive ability and text genre (narrative vs. expository texts) and text properties (emotional vs. scientific texts). However, previous neuroimaging work has mostly focused on the comprehension of narrative, story-based texts in native language (L1) readers. In this study we examined the neurocognitive mechanisms of comprehension of science texts in both L1 and L2 (second language) readers. We aim at identifying how students acquire (or fail to understand) scientific concepts in the STEM domains (Science, Technology, Engineering, and Mathematics). Our findings indicate that (1) reading comprehension and the individual’s reading ability were correlated. High- skilled readers showed stronger connectivity among brain regions important for integrating semantic representation and building event representation; (2) graph centrality (GC), a measure of knowledge structure, showed positive effects on lexical- and sentence-level comprehension in angular gyrus and dorsolateral prefrontal cortex (dlPFC), regions for semantic and conceptual integration and working memory. Group-level regression with behavioral measures also showed that for texts with higher GC, readers’ e-device time per day is negatively correlated with their fMRI activity in the right dlPFC and cerebellum, suggesting that electronic habits might affect readers’ ability to integrate conceptual representation in a structured way; and (3) in a Representation Similarity Analysis whole-brain voxel-wise activities for all pairs of words in the texts were correlated with computational models based on visual or semantic features, and semantic models predicted brain activity better than visual models but only for L1 readers, suggesting that L2 proficiency impacts the quality of the conceptual representation established. These findings provide initial findings into the neurocognitive underpinnings of science text comprehension.