Cortico-Cortical Feedback

 描述(由申请人提供)：皮质-皮质反馈联系的普遍存在有力地证明了它们的重要性，理论家经常在皮质功能的模型中使用它们。尽管如此，人们对它们的功能几乎一无所知。之前的研究在 我们在麻醉动物身上进行了哪些反馈被操纵，我们最近对警觉动物的研究揭示了根本不同的结果。我们在前一个周期的主要发现是，来自V2的反馈对V1神经元经典定义的感受野中心的影响很小：方向调整和中心的对比度反应功能几乎没有受到影响。然而，当V2失活时，较大的刺激与周围的感受野接触并通常会导致抑制，导致的抑制要小得多。我们进一步发现，反馈的影响是对环绕的空间范围，而不是对其增益，以及似乎更高阶的环绕抑制，如特征匹配抑制，更强烈地受到反馈的影响，而不是基本的归一化。所有这些观察结果都与“预测编码”的理论框架一致，该框架假定前馈和反馈连接之间的相互作用具有关键的计算作用。我们现在建议考察具体的例子，从形式和运动知觉的内在语境影响到作为知觉任务学习基础的变化的性质。我们的一般方法将是使用因果操作-反馈到V1的皮质区域(V2和/或MT)的可逆失活-来测试自上而下影响在知觉推理中的假设作用。我们的研究将有助于揭示大脑皮质功能的基本机制。

 DESCRIPTION (provided by applicant): The ubiquity of cortico-cortical feedback connections argues strongly for their importance, and theoreticians routinely make use of them in models of cortical function. Despite this, next to nothing is known about their function. Previous studies in which feedback has been manipulated have been performed in anesthetized animals, and our recent studies in alert animals have revealed fundamentally different results. Our major finding from the previous cycle is that feedback from V2 exerts minimal effects on the classically defined receptive field center of V1 neurons: orientation tuning and the center's contrast response function are virtually unaffected. However, larger stimuli that engage the receptive field surround and that normally cause suppression, cause much less suppression when V2 is inactivated. We have further discovered that the influence of feedback is on the spatial extent of the surround, rather than on its gain, and that seemingly higher order properties of surround suppression, such as feature- matched suppression, are more strongly influenced by feedback than is basic normalization. All of these observations are consistent with a theoretical framework of "predictive coding," which posits a key computational role for the interaction between feedforward and feedback connections. We now propose to examine specific examples ranging from built-in contextual influences in form and motion perception to the nature of changes that underlie the learning of a perceptual task. Our general approach will be to use causal manipulations-reversible inactivation of cortical areas (V2 and/or MT) that feed back to V1-to test the hypothesized role of top-down influences in perceptual inference. Our studies will shed light on the basic mechanisms of cortical function.