Color Constancy for an Unseen Surface

Abstract

The illumination of a scene strongly affects our perception of objects in that scene, e.g., the pages of a book illuminated by candlelight will appear quite yellow relative to other types of artificial illuminants. Yet at the same time, the reader still judges the pages as white, their surface color unaffected by the interplay of paper and illuminant. It has been shown empirically that we can indeed report two quite different interpretations of "color": one is dependent on the constant surface spectral reflectance of an object (surface color) and the other on the power of light of different wavelengths reflected from that object (reflected color). How then are these two representations related? The common view, dating from Aristotle, is that our experience of surface color is derived from reflected color or, in more familiar terms, that color perception follows from color sensation. By definition, color constancy requires that vision "discounts the illuminant"; thus, it seems reasonable that vision begins with the color of objects as they naively appear and that we infer from their appearances their surface color. Here, we question this classic view. We use metacontrast-masked priming and, by presenting the unseen prime and the visible mask under different illuminants, dissociate two ways in which the prime matched the mask: in surface color or in reflected color. We find that priming of the mask occurs when it matches the prime in surface color, not reflected color. It follows that color perception can arise without prior color sensation.