Contextual Predictions Facilitate Visual Cognition

描述（由申请人提供）：我们环境中的对象倾向于在典型的上下文中进行分组。我们假设，人类的大脑可以快速提取这些背景信息，并利用它来产生预测，从而有效地指导认知。在一系列的行为和神经成像实验中，我们之前已经揭示了在日常物品的语境处理过程中唯一被激活的皮层网络。该网络由三个主要病灶组成，分别位于海马旁皮层、脾后皮层和眶上前额叶皮层。在初步结果的基础上，我们提出了15项实验，以实现三个具体目标。在目标1中，我们将定义调节上下文处理的皮层网络中每个主要组成部分的相对作用和特性，并描述其编排背后的皮层动力学。在目标2中，我们将研究上下文处理和对象识别之间的相互作用，并将测试我们的上下文如何促进识别的模型。具体来说，我们建议基于图像中的粗糙信息快速提取上下文信息，随后得出有助于视觉对象识别的预测。最后，Aim 3旨在测试我们的建议，即上下文表示包括关于倾向于共享相同“上下文框架”的对象的主要信息，以及关于它们之间典型空间关系的信息。实验结合了功能磁共振成像（fMRI）、脑磁图（MEG）、弥散张量成像（DTI）、计算方法和纯粹的认知和心理物理实验。场景识别和上下文关联对于我们日常生活的许多方面都是至关重要的，理解它们是如何完成的对于任何完整的大脑理论都是必不可少的。最后，虽然拟议研究的重点是基础研究，但预计这些机制的表征将为评估精神疾病期间的功能障碍提供一个平台。

DESCRIPTION (provided by applicant): Objects in our environment tend to be grouped in typical contexts. We hypothesize that the human brain extracts such contextual information rapidly and uses it to generate predictions that subsequently guide cognition efficiently. In a series of behavioral and neuroimaging experiments, we have previously revealed the cortical network that is uniquely activated during contextual processing of everyday objects. This network is comprised of three primary foci, located in the parahippocampal cortex, retrosplenial cortex and superior orbital prefrontal cortex. Building on our initial results, we propose fifteen experiments to address three specific aims. In Aim 1 we will define the relative role and properties of each major component in the cortical network that mediates contextual processing, and characterize the cortical dynamics underlying their orchestration. In Aim 2 we will study the interaction between contextual processing and object recognition, and will test our model of how context facilitates recognition. Specifically, we propose that contextual information is extracted rapidly based on coarse information in the image, which subsequently derives predictions that facilitate visual object recognition. Finally, Aim 3 is designed to test our proposal that contextual representations include gist information about objects that tend to share the same "context frame," as well as information about the typical spatial relations among them. The experiments integrate functional magnetic resonance imaging (fMRI), magnetoencephalography (MEG), diffusion tensor imaging (DTI), computational methods and purely cognitive and psychophysical experiments. Scene recognition and contextual associations are fundamentally important for many aspects of our day-to-day lives, and understanding how they are accomplished is essential for any complete theory of the brain. Finally, while the focus of the proposed studies is on basic research, it is anticipated that the characterization of these mechanisms would provide a platform from which to evaluate their dysfunctions during mental illness.