Functional Neuroimaging of Face and Object Representations in the Ventral Visual Pathway

对人脸和其他物体类别的视觉识别在人类腹侧视觉通路的纹状外皮层唤起不同的、与类别相关的反应模式。在NSF的资助下，詹姆斯·V·哈克斯比博士正在研究人类大脑腹侧视觉通路中不同面孔的类别内差异的表征。此前，他和他的同事们证明，次最大神经反应的模式也携带与类别相关的信息，这表明不同对象类别的表征是重叠的，弱和强神经反应都是代表对象外观的群体反应所不可或缺的。然而，以往的研究只讨论了不同客体类别表征之间的差异，只分析了个体内部反应模式的一致性，而没有分析个体之间反应模式的一致性。这项受资助的研究的目的是调查单一类别中样本之间的区别，即人脸。实验正在测试面孔之间类别内差异的表征是否在腹侧通路中有更集中的解剖分布，即类别内差异是在对面孔反应最大的区域表现出来的，还是在更广泛分布的区域表现出来的。研究内容包括：(1)检验神经反应对类别内刺激变化的分布；(2)分析刺激引起的反应模式。面孔外观是沿着两个计算定义的维度进行操作的，一个维度反映了与性别相关的外观的连续变化，另一个维度反映了从瘦到宽的连续变化，以调查神经反应模式和计算定义的面容变化之间的相关性。与其他研究人员一起，除了使用地形模式分析方法外，正在开发更复杂的方法(如主成分分析、独立成分分析)来解剖反应模式，以识别携带不同方面信息的部分或子空间。这些研究代表了一种新的方法来阐明腹侧视觉通路中面孔表征的详细结构。其长期目标是发现构成面部表征拓扑图的组织原理，从而揭示神经反应模式与面部外观信息的计算描述之间的对应规则。这些按地形组织的人口反应的组织原则也可能与研究其他类型的信息有关，如视觉运动、听觉和语言。开发新的方法来分析神经成像数据中的反应模式将使整个功能神经成像领域受益，并将促进对抽象信息的神经表示的新视角的发展。使用计算模型来指导面孔神经表征的研究将促进对面孔如何被识别，社会刻板印象和种族等因素如何影响面孔识别，以及面孔感知在社会交流中的作用的理解。这项研究涉及一名博士后研究员和研究生的努力。它正在帮助普林斯顿大学建立认知神经科学的研究和教育项目。这项研究正在普林斯顿大学的课堂上使用，并将作为指导学生研究项目的基础，并帮助来自其他领域的研究人员，如社会心理学，将功能神经成像应用于相关问题。鼓励来自代表性不足群体的学生参与研究团队并完成研究项目。与计算模型师和应用数学家的合作促进了心理学家学习新的计算方法，并有助于点燃其他学科对认知神经科学的兴趣。

Visual recognition of faces and other object categories evokes distinct, category-related patterns of response in the extrastriate cortices of the human ventral visual pathway. With funding from NSF, Dr. James V. Haxby is investigating the representations of within-category distinctions among faces in the ventral visual pathway of the human brain. Previously, he and his colleagues showed that the patterns of submaximal neural responses also carry category-related information, indicating that the representations of different object categories are overlapping, and that weak and strong neural responses are both integral to population responses that represent object appearance. However, previous studies only addressed the distinctions between representations for different object categories and only analyzed the consistency of patterns of response within but not across individual subjects. The objective of the funded research is to investigate the representation of distinctions between exemplars within a single category, namely human faces. Experiments are testing whether representations of within-category distinctions among faces have a more focal anatomical distribution in the ventral pathway, that is, whether within-category differences are represented in regions that respond maximally to faces or are more broadly distributed. Studies are (1) examining the distribution of neural responses to within-category stimulus changes, and (2) analyzing the patterns of response evoked by stimuli. Face appearance is manipulated along two computationally-defined dimensions, one that reflects continuous variation in gender-related appearance and a second that reflects continuous variation from lean to wide, to investigate the correlation between patterns of neural response and computationally-defined alterations of face appearance. With other investigators, in addition to using methods of topographic pattern analysis, more sophisticated methods (e.g. PCA, ICA) are being developed for dissecting patterns of response that can identify the parts or subspaces that carry different aspects of information. These studies represent a novel approach to elucidating the detailed structure of the representation of faces in the ventral visual pathway. The long-term goal is to discover the principles of organization that underlie the topography of face representations and, thereby, reveal the correspondence rules that relate patterns of neural response to computational descriptions of information about face appearance. These principles of organization for topographically-organized population responses also may be relevant for the study of other types of information, such as visual motion, audition, and language. Development of new methods for analyzing patterns of response in neuroimaging data will benefit the entire field of functional neuroimaging and will facilitate the development of a new perspective on the neural representation of abstract information. Using computational models to guide the research on neural representation of faces will advance understanding of how faces are recognized, how factors such as social stereotypes and race affect face recognition, and the role face perception plays in social communication.Broader impacts. This research involves the efforts of a postdoctoral fellow and graduate students. It is helping to build the research and education program in cognitive neuroscience at Princeton University. The research is being used in classes at Princeton and will serve as the basis for guiding student research projects and for helping researchers from other fields, such as social psychology, apply functional neuroimaging to related questions. Students from under-represented groups are encouraged to become involved with the research team and to complete research projects. Collaborations with computational modelers and applied mathematicians facilitate the learning of new computational methods by psychologists and serve to ignite interest in cognitive neuroscience in other disciplines.