Visual memory system and mechanisms of associative learning : Interaction between the temporal neocortex and the hippocampus

神经心理学理论提出内侧颞叶和新皮层之间的相互作用在形成对事实和事件的长期记忆中起着关键作用，这通常通过人类学习一系列成对的单词或数字来测试。在本研究中，通过对猴子进行视觉配对联想任务的动物模型进行单单位记录，对记忆形成的神经机制进行了研究。首先，验证了视觉对象的长时表征是通过前下颞叶（IT）皮层的神经网络学习获得的假说。我确定了两种机制;一种是调优，另一种是关联。在执行视觉配对联想任务的猴子的IT皮层中，我发现一组神经元对两个配对联想都表现出选择性反应（配对编码神经元）。这为单个IT神经元通过联想学习获得刺激选择性提供了有力的证据。 ...更多信息 宁然后，我通过询问反复暴露于图片是否会产生IT神经元来测试调谐机制，这些神经元选择性地对特定形式做出反应。单神经元的活动进行了比较几何相似的模式，只有其中一个已被反复暴露给猴子之前的单位记录（学习模式）。我发现IT神经元对学习的模式反应最好。对学习模式的反应是最优的，然而，只是局部的，并且在单个神经元的整个形式空间中可能有几个局部最大值。一个配对编码神经元在形式空间中至少有两个独立的局部极大值，对应于每一个配对的联想。第二，我测试了从内侧颞叶到IT皮层的反向信号的作用。将鹅膏蕈氨酸单侧注射到内嗅和嗅周皮质中，所述内嗅和嗅周皮质向同侧IT皮质提供大量向后投射。损伤后，猴子可以学习一组新的配对联想。我发现，（i）尽管有损伤，神经元对视觉刺激有选择性的反应，这表明调谐机制的保留，（ii）成对的联想未能在损伤的细胞中引起相关反应。我的结论是，边缘系统的病变破坏了配对的联营公司之间的IT神经元的关联代码，而不损害对每个刺激的视觉反应。这些结果为灵长类动物的长期记忆系统分解为不同的神经机制提供了具体的证据。少

Neuropsychological theories proposed a critical role of the interaction between the medial temporal lobe and neocortex in the formation of long-term memory for facts and events, which has often been tested by learning of a series of paired words or figures in humans. In this research project, I have been identifying neural mechanisms underlying the memory formation by single-unit recording with an animal model of visual pair-association task in monkeys.First, I tested a hypothesis that long-term representations of visual objects are acquired through learning in the neural network of the anterior inferotemporal (IT) cortex. I identified two mechanisms ; one is tuning and the other is association. In the IT cortex of monkeys performing the visual pair-association task, I found a group of neurons that manifested selective responses to both of the paired associates (pair-coding neuron). It provides strong evidence that single IT neurons acquire stimulus-selectivity through associative lear … More ning. Then I tested tuning mechanism by asking whether repeated exposures to a picture produce IT neurons that selectively respond to that particular form. Single-neuron activity was compared among geometrically similar patterns, only one of which had been exposed repeatedly to the monkey before the unit recording (learned pattern). I found that IT neurons responded best to the learned pattern. The reponse to a learned pattern is optimal, however, only locally, and there are presumably several local maxima in the whole form space of a single neuron. A pair-coding neuron would have at least two separate local maxima, correponding to each of the paired associates, in the form space.Second, I tested the role of the backward signal from medial temporal lobe to IT cortex. Ibotenic acid was injected unilaterally into the entorhinal and perirhinal cortex which provide massive backward projections ipsilaterally to IT cortex. After the lesion, the monkeys could learn a new set of paired associates. I found that (i) in spite of the lesion, the neurons responded selectively to the visual stimuli, which suggests preservation of the tuning mechanism, and (ii) the paired associates failed to elicit correlated responses in the cells tested with the lesion. I conclude that the limbic lesion disrupted the associative code of the IT neurons between the paired associates, without impairing the visual response to each stimulus. These results give concrete evidence for decomposition of the primate long-term memory system into different neural mechanisms. Less