Abstract

Human beings are among the species with the best color perception of all mammals. Yet, transparency can be perceived in scenes in which color cues point to opacity. Why do we ignore such color cues? Here we argue that colors, rather than being passively registered, must be actively recreated and then bound to other stimulus attributes. In this process, the visual system faces fundamental problems, some of which are logically impossible to solve. The resulting unreliability of color perception may go some way toward explaining why color cues cannot usually veto transparency percepts. Other stimulus attributes, however, affect transparency perception strongly even though they are not processed in foolproof ways either. How come? We argue that our trichromatic color perception is likely to have been a late, and for that reason less than fully integrated, addition to an already existing visual system that was colorblind to red and green.