Prediction in Visual Cortex

描述（由申请人提供）：大脑最基本和最一般的功能之一是检测意外事件的发生，因此可能提供信息。这个功能建立在两个基础上。首先，大脑必须从持续的经验中提取关于环境过渡统计的信息：关于什么事件倾向于跟随什么事件的信息。其次，它必须监测正在进行的经验，以便在基于以前经验的过渡统计数据的预测被违反时发出信号。我们知道大脑完成了这些壮举，但我们对潜在的神经元机制知之甚少。我们建议通过研究灵长类视觉系统中最近发现的一种现象：预测抑制来研究这个问题。预测抑制是通过让猴子反复观看一个固定顺序的画面来诱导的。在随后的观看过程中，颞叶视觉区域TE的神经元对训练序列中的图像反应较弱，但对出现在序列之外的图像反应强烈。这一现象具有作为研究大脑中调节预测过程的神经机制的试验台的潜力。该项目的总体目标是回答关于预测抑制的基本未决问题，从而巩固该现象作为研究预测过程神经机制的模型的地位。我们提出了三组实验。在所有这些实验中，我们将在训练期间让猴子观看由固定的过渡统计数据控制的动态显示，然后在测试期间，我们将监测神经元对预测和未预测序列的反应。第一个系列的实验将描述诱导预测抑制所需的条件。第二个系列的实验将描述由预测刺激引发的神经过程。第三系列的实验将确定在区域TE水平之前和之后的视觉加工阶段是否存在预测抑制。

DESCRIPTION (provided by applicant): One of the most basic and general functions of the brain is to detect the occurrence of unexpected and therefore potentially informative events. This function rests on two foundations. First, the brain must extract from ongoing experience information about the transitional statistics of the environment: information about what events tend to follow what other events. Second, it must monitor ongoing experience so as to signal when predictions based on previously experienced transitional statistics are violated. We know that the brain accomplishes these feats but we know little about the underlying neuronal mechanisms. We propose to investigate this problem by studying a phenomenon recently discovered in the primate visual system: prediction suppression. Prediction suppression is induced by allowing a monkey to view repeatedly a display in which images follow each other in fixed sequence. During subsequent viewing, neurons in area TE, a visual area in the temporal lobe, respond weakly to images in the trained sequence but respond strongly to images that appear out of sequence. This phenomenon possesses potential as a test bed for the study of neural mechanisms mediating predictive processes in the brain. The general goal of the proposed project is to answer fundamental outstanding questions about prediction suppression so as to solidify the status of the phenomenon as a model for studying the neural mechanisms of predictive processes. We propose three sets of experiments. In all of them, we will expose monkeys, during a training period, to dynamic displays governed by fixed transitional statistics and then, during a testing period, will monitor neuronal responses to predicted and unpredicted sequences. Experiments in the first series will characterize the conditions required for inducing prediction suppression. Experiments in the second series will characterize the neural processes initiated by presenting a predictive stimulus. Experiments in the third series will determine whether prediction suppression is manifest at visual processing stages before and after the level of area TE.