Functional Roles of the Top-Down Information Processing in Cognitive Processes

为了阐明自上而下加工的功能作用，我们进行了一系列的生理和心理实验，沿着数学分析。采用全细胞膜片钳技术记录鲤鱼视网膜细胞对光刺激的反应。这种制剂被发现是有用的研究视网膜中的反馈机制。选择性注意的神经元相关性是从记录猴子前额叶皮层单个单位活动和场电位的实验中获得的。认知研究表明，视觉突然变化引起的选择性注意的时间特征表现在对视运动方向的判断上。在运动后效的心理物理学实验中，发现需要选择注意的并发任务并不减少复杂运动（如伸展、收缩和旋转）的运动后效持续时间，而减少简单平移的运动后效持续时间。本课题的心理学实验还探讨了人类色觉中的黄蓝机制、日常物体的色觉、绝对音高拥有者的音高知觉以及人类概率判断的偏差等问题。在对人类信息处理机制的理论探讨中，建立并分析了数学模型;在对神经网络模型分析中提出了一种新的学习方法;建立了人类狭缝视觉的计算模型;在分析特征检测与计算复杂度关系的基础上，提出了计算理论。

In order to elucidate the functional roles of the top-down processing, we conducted a series of physiological and psychological experiments along with mathematical analyzes. Responses of retinal cells to light stimulus were recorded in carp slice preparation using the whole-cell patch-clamp technique. Such preparations are found to be useful for studying feedback mechanisms in retina. Neuronal correlates of selective attention were obtained from the experiments in which single unit activity and field potential were recorded in the monkey prefrontal cortex. From the cognitive studies, temporal characteristics of the selective attention induced by an abrupt change in the visual field were featured in terms of the decision of the direction of apparent motion. In the psychophysical experiment of motion aftereffect, it was found that a concurrent task which required selective attention did not reduce the duration of the motion aftereffect of complex motion, such as expansion, contraction and rotation, though it did reduce that of simple translation. Other problems investigated in the psychological experiments of our proiect were as follows : yellow-blue mechanism in human color vision, color perception for daily objects, pitch perception of absolute pitch possessors and bias in probability judgment by humans. In the theoretical approach to the foundations of the mechanisms of human information processing, mathematical models were formulated and analyzed ; a new learning method was proposed in the analysis of neural network models, a computation model was formulated for human slit vision and a computation theory was proposed on the basis of the analysis of the relationship between feature detection and computational complexity.