A Neural Network Model for Categorical Effects in Color Memory

Human color perception is categorical. Previous experimental studies have shown that the color category has profound effects on cortical neural responses, perceptual color discrimination and color memory. However, existing theoretical studies are not enough to provide an inclusive model accounting for those categorical effects in color perception and memory. In this study, we propose a computational model for categorical color processing, where the color memory is represented by a population of color selective neurons in cortex Our model reproduces the characteristics of color memory reported in the previous experimental studies. Furthermore, it explains the properties of neurons in IT cortex, which change the activity depending on whether the task demands is color categorization or discrimination. This study suggests that perceptual biases found in color processing and task-dependent modulations of neural responses may be explained as a natural consequence of statistically optimal estimation.