The mechanism of mental scanning in foveal vision
The mechanism of mental scanning in foveal vision
复制标题
中心凹视觉的心理扫描机制
DOI：
10.1007/bf00337143
复制
发表时间：
1978
影响因子：
1.9
通讯作者：
R. Suzuki
中科院分区：
文献类型：
--
作者：
T. Inui;M. Kawato;R. Suzuki
为了感知包含多个基本图片的视觉模式，感知者必须将其认知资源分配给适当选择的模式部分并按顺序扫描它们。即使视野很小并且不需要眼球运动，也能发现这种扫描。我们称之为“心理扫描”，并进行了心理实验来研究其机制。任务是辨别图案中的元素图片是否全部相同（SP）或不同（DP）。在不同的暴露持续时间下测量了任务的正确百分比。我们将阈值定义为获得 75% 正确答案的暴露持续时间。我们的主要发现如下。 SP的阈值与图像元素的数量成正比，而DP的阈值是恒定的。似乎存在两种心理扫描模式。一种是SP的串行处理，一种是DP的并行处理。我们提出了具有以下特征的两层神经网络模型。 1）信息通过两个独立的通道作为两种类型的信号传输；一种是传输到Y层的瞬态信号，另一种是缓慢传导到X层的持续信号。 2）Y层神经元之间的相互作用是侧向抑制性的，而X层神经元之间的相互作用是自兴奋性和侧向抑制性的。 3）Y层中的每个神经元向X层中的每个神经元发送抑制信号，除了具有相同感受野的神经元。在这些条件下，X层神经元的动力学由一组特定方程表示。从相平面分析和数值积分来看，该模型似乎能够解释各种实验结果。
In order to perceive a visual pattern which includes several elemental pictures, the perceiver must allot his cognitive resources to suitably selected parts of the pattern and scan them in sequence. Even when the visual field is small and eye-movement is not required, such scanning is found. We called it ‘mental scanning’ and performed psychological experiments to investigate the mechanism. The tasks were to discern whether the elemental pictures in a pattern are all the same (SP) or not (DP). The per cents correct of the task were measured for various exposure durations. We defined the threshold as the exposure duration at which 75% correct answers were obtained. Our main findings are as follows. The threshold for SP is proportional to the number of picture elements, while the threshold for DP is constant. It appears that two modes of mental scanning exist. One is serial processing for SP, and the other is parallel processing for DP. We proposed a two-layered neural network model having the following characteristics. 1) Information is transmitted as two types of signals through two separate channels; one is the transient signals to the Y layer and the other is the sustained signals slowly conducted to the X layer. 2) Interactions among neurons in the Y layer are lateral inhibitory, while those in the X layer are self-excitatory and lateralinhibitory. 3) Every neuron in the Y layer sends inhibitory signals to every neuron in the X layer except one with the same receptive field. Under these conditions, the dynamics of neurons in the X layer is represented by a set of certain equations. From phase plane analysis and numerical integration, the model appears to have an ability to account for various experimental results.