Two Kinds of Adaptation Model for General Perceptual Dynamics in Binocular Rivalry

When two different images are presented to the left and right eyes simultaneously, only one stimulus is perceived, and its percept is intrinsically alternated. This psychological phenomenon is known as binocular rivalry. Previous studies on binocular rivalry revealed many characteristics as summarized as follows: (1) Levelt's propositions, (2) Mismatch effect, and (3) Flash effect. However, not a single neural network model can explain all of them yet. The purpose of this paper is to construct a neural network model that can explain these three experimental characteristics. Hypothesizing that two kinds of adaptation, slow and predictive adaptations are important, we constructed a model including both of them. Then we compared the computer simulation results with three neural network models: the slow adaptation model, the predictive adaptation model, and the proposed model. The results showed that only the proposed model satisfied all of the characteristics. This suggests that both slow and predictive adaptations are necessary to explain the phenomena in binocular rivalry.