Visual processing is not static but is influenced by context, enabling vision to rapidly adjust to the varying conditions in natural viewing. In color vision, contextual interactions are thought to support efficient information processing and perceptual tasks like color constancy. We investigated context-induced perceptual biases in color vision, as well as internal biases derived from discrimination experiments with noisy color stimuli using a Bayesian observer model.
Context-induced biases showed a systematic pattern consistent with effects of modulatory interactions in a population code. The bias pattern was similar to those for other features like orientation or motion perception, indicating that common neural mechanisms underlie sensory processing across visual domains. Bias magnitudes were not uniform in color space but were stronger for colors corresponding to natural daylights. Similarly, based on the experimental data the Bayesian observer model predicted a prior favoring bluish colors. The findings indicate that color vision is adapted to the regularities of the sensory environment.