NEURAL CODING OF VISUAL STIMULI

视觉系统内不同区域的神经元执行 视觉所需的视觉模式分析的不同方面 感知力。我们正在研究如何以及在哪里组成一个 视觉感知的对象被聚集在一起。我们录下了 初级视皮层单个神经元的活动(第一 视觉处理的皮质阶段)，以及颞叶下部皮质( 大脑皮层的最后阶段)来研究视觉的潜在机制 感知力。当我们问屏幕上的眼球运动是如何发生的 会影响初级视觉皮质神经元的反应 发现即使眼睛需要300-400毫秒才能完全稳定下来 (包括取决于时间结构的消息部分 的反应)，与刺激相关的信息量在 这些反应几乎与他们在 稳定下来了。这种对视网膜位置不稳定的容忍 视觉刺激提示初级视觉皮质神经元对 刺激内特征的整体结构，而不是 自身的特点。神经元开始放电前的潜伏期 出现的模式是时间结构的一个重要方面 神经信息。我们发现，延迟时间表明 模式可以看到，射击的强度，反应 力量，是一种指示模式是什么的代码。我发现， 潜伏期与观察模式的难易程度有关 阵列中的特征被分组。我们构建了一个模型来 显示神经元信号如何从外侧进入视觉皮质 膝状核可以结合在一起，使这种特征组合发生。 当下颞叶皮质神经元对一种模式做出反应时 视觉噪音，猴子对图案做出反应所需的时间 正确的预测是通过神经元所需的时间 指示存在哪种模式的消息。我们在两端的结果 从视觉皮层流来看有一个模板 构成了一种有效的刺激模式，它可以控制神经元的反应。 所有这些发现加在一起表明，视觉系统 发送将刺激描述为 从最早的皮质阶段开始的复杂、完整的物体，以及 当信号传递到以后时，这种积分变得更普遍 视觉处理的各个阶段。

Neurons in different regions within the visual system carry out different aspects of visual pattern analysis needed for visual perception. We are studying how and where the pieces that make up a visually perceived object are brought together. We recorded the activity of single neurons in the primary visual cortex (the first cortical stage of visual processing), and inferior temporal cortex (the last cortical stage) to study the mechanisms underlying visual perception. When we ask how the eye movements that occur on a screenful of images affect the responses of primary visual cortical neurons we find that even though the eyes take 300-400 ms to stabilize completely (including the parts of the message depending on the temporal structure of the response), the amount of stimulus-related information found in these responses is almost as much as that found after they have stabilized. This toleration of instability in the retinal location of a visual stimulus suggests that primary visual cortical neurons react to the overall structure of the features within a stimulus, not the features themselves. The latency before neurons start firing after a pattern appears is one important aspect of the time structure of a neuronal message. We find that the latency indicates how easily the pattern can be seen, and the intensity of the firing, the response strength, is a code indicating what the pattern is. Finding that the latency is related to how easily the pattern can be seen suggests that the features within in a pattern are grouped. We construct a model to show how the neuronal signals entering visual cortex from the lateral geniculate nucleus can be combined to make this feature grouping occur. When inferior temporal cortical neurons respond to a pattern obscured by visual noise, the time it takes for the monkey to respond to the pattern correctly is closely predicted by the time it takes for the neuronal messages to indicate which pattern is present. Our results at both ends of the visual cortical streams suggests there is a template of what constitutes a valid stimulus pattern that gates the neuronal responses. All of these findings taken together suggest that the visual system is sending messages that describe messages that describe stimuli as complex, integrated objects from the earliest cortical stages, and that this integration becomes more general as the signals pass to later stages of visual processing.