Abstract
The notion of unconscious perceptual inference, particularly in the shape of predictive coding, promises to address many of the problems of visual perception that has long preoccupied philosophy. The dichoptic viewing conditions during binocular rivalry presents an extreme example of the epistemic task confronting perceptual inference and rivalry is therefore an important test-case for the predictive coding framework. This talk explores what predictive coding can and cannot tell us about the mechanism behind rivalry. The guiding thought is that Bayesian inference and prediction error drive selection, dominance and alternation in rivalry, and a number of computational and empirical studies of rivalry are reviewed in this light. The talk also presents preliminary findings from interdisciplinary imaging and psychophysical studies of binocular rivalry, which test aspects of the predictive coding hypothesis.