Visual Adaptation and Neuronal Selectivity

在灵长类动物的大脑中，面部细胞，或者更普遍的复杂图案细胞的存在是在灵长类大脑中最引人注目的发现之一。无论是名人的视觉刺激还是一种食物的视觉刺激，对特定视觉刺激的反应高度特异的神经元在这一区域都很丰富。虽然这些神经元的存在显然与识别和刺激记忆有关，大量的损伤实验证明了这一点，但它们也位于复杂的一般处理流的末端，其生理原理尚未完全阐明。也许出于这个原因，它们在视觉感知、图片记忆和更一般的物体识别中对特征或物体反应神经元的确切作用一直很难确定。 我们最近的工作详细阐述了细胞如何针对人脸进行调整，以促进对个人的识别。简而言之，我们发现，颞下皮质中的神经元不是根据构成面孔的绝对特征做出反应，而是根据给定面孔偏离所有面孔的平均值的程度做出反应。心理物理实验表明，我们对内隐存储的原型的面孔处理是不同的。而颞下皮质中的神经元似乎直接促成了这种复杂的分析模式。 一个基本的问题是，这种神经选择性是如何产生的？神经元是否与生俱来地专用于某些类型的刺激？它们是由经验塑造的吗？如果是这样的话，他们的任务有多长时间？虽然已经有许多理论阐述了这些和类似的问题，但经验证据相对较少。我们之前已经使用了一种技术，允许我们一次分离和监测单个神经元数天和数周。在这些研究中，猴子被一遍又一遍地展示一系列刺激，而不需要任何特定的视觉学习，我们发现，随着时间的推移，下颞叶皮质的神经反应非常稳定，不仅保持了它们的总体选择性，而且保持了它们精确的时间模式。 在实验室，我们正在扩展这项工作，在旨在了解神经选择性是如何产生并随着时间的推移而形成的研究的背景下。当一个人遇到一群新人，或者突然养成了一种需要视觉特化的爱好(例如观鸟)时，下颞叶皮质会发生什么？这些问题是触手可及的，因为我们可以通过各种方式训练猴子成为视觉专家。例如，出现的一个问题是，当学习新面孔时，曾经专用于特定一张或多张面孔的神经元是否会改变其选择性。一个相关的问题是，神经反应是“向后兼容的”--如果它们开始对新的刺激做出反应，它们是否会继续对它们以前喜欢的刺激做出反应？ 我们的研究将在以规范为基础的理论背景下进行。本文认为，视觉专门化从根本上关注的是分析与“原型”刺激的偏差。以人脸为例，给定的人脸不是根据其精确的度量标准编码的，而是通过它与原型的偏差来编码的，可能是所有人脸的平均值。这是一个微妙的观点，但它对视觉和其他感觉处理可能如何进行有着重要的影响。这也得到了最近的研究的支持，该研究表明，这一区域的神经元围绕平均面部进行调节，这在整个人群中是特殊的。我们目前的问题取决于我们是否能够在基于原型的方案的背景下通过改变学习压力来塑造神经调谐。如果是这样的话，调谐变化的性质将非常有助于了解复杂的刺激信息在大脑这一区域是如何编码的。 最后，我们正在研究几秒钟对一种刺激(直线、曲线和脸部)的适应如何在知觉上“立即”留下印记。几十年前的研究结果表明，仔细观察自己的感觉的观察者很容易察觉到这种适应(例如，一条曲线自发地看起来略微变直了)。这些发现提供了直接的证据，表明我们的知觉主动地“倾向”于特定的刺激，也许可以被认为是吸引器刺激的原型。该项目这方面的长期目标是将这种主观适应与视觉皮质不同处理阶段单个神经元的放电率的良好记录的适应联系起来。

The existence of face cells, or more generally complex pattern cells, in the inferotemporal cortex is one of the most striking findings made in the primate brain. Neurons that are highly specific in their responses to a particular visual stimulus, whether that of a celebrity or a type of food, are abundant in this area. While the presence of these neurons is clearly related to recognition and stimulus memory, as evidenced by a large body of lesion experiments, they also sit at the end of a complex general processing stream whose physiological principles have yet to be completely elaborated. Perhaps for this reason, their precise role of such feature- or object-responsive neurons in visual perception, picture memory, and more general object recognition has been difficult to ascertain. Our recent work has elaborated how cells may be tuned for faces in a way that facilitates the recognition of individuals. Briefly, we found that neurons in inferotemporal cortex respond not according to the absolute features that constitute a face, but rather in the extent to which a given face deviates from the average of all faces. Psychophysical experiments have shown that we process faces differentially with respect to an implicitly stored prototype. And neurons in the inferotemporal cortex appear to contribute directly to this mode of complex analysis. A fundamental question is, how does such neural selectivity come about? Are neurons innately dedicated to certain types of stimuli? Are they shaped by experience? And, if so, how permanent are their assignments? While numerous theories have been elaborated that address these and similar questions, there is relatively little empirical evidence. We have previously used a technique that permits us to isolate and monitor single neurons for days and weeks at a time. In these studies, in which monkeys were shown a set of stimuli over and over again, without any particular visual learning required, we found that neural responses in the inferotemporal cortex were remarkably stable over time, maintaining not only their general selectivity, but also their precise temporal patterning. In the laboratory, we are extending this work, in the context of studies aiming to understand how neural selectivity comes about and is shaped over time. What happens in the inferotemporal cortex when one meets a group of new people, or suddenly takes up a hobby that requires visual specialization (e.g. birdwatching)? These questions are within reach, as we can train monkeys to become visual specialists in a variety of ways. One question that arises, for example, is whether neurons, once dedicated for a particular face or faces, change their selectivity when new faces are learned. A related question is, are neural responses "back-compatible" -- if they start responding to a new stimulus, do they continue to respond to those which they have previously favored? Our research will be conducted in the context of a norm-based theoretical context. This context considers that visual specialization is fundamentally concerned with analyzing deviations from a "prototype" stimulus. To use the example of faces, a given face is coded not by its precise metrics, but rather by its deviation from a prototype, perhaps the average of all faces. This is a subtle point, but it has important implications for how visual and other sensory processing might be done. It is also supported by recent work, which shows that neurons in this area are tuned around the average face, which is special across the population. Our present question hinges on whether we are able to shape neural tuning by changing the learning pressures in the context of a prototype-based scheme. If so, the nature of the tuning changes will be highly informative about how complex stimulus information is encoded in this region of the brain. Finally, we are investigating how the adaptation to one stimulus for several seconds (lines, curves, and faces) makes an "immediate" mark on perception. Decades-old findings show that such adaptation is readily detectable by an observer attending carefully to her/his own perception (e.g. a curved line spontaneously appears to straighten ever so slightly). Such findings provide direct evidence that our perception actively "tends" toward particular stimuli, perhaps to be thought of as a prototype attractor stimulus. The long term goal of this aspect of the project is to link such subjective adaptation to the well-documented adaptation in the firing rate of individual neurons at different stages of processing in the visual cortex.