By hypothesis we state that there exists a set of three color channels, linear combinations of the cones, for which a simple von Kries type scaling, or coefficient rule, is an adequate vehicle for color constancy. We develop a computational method for solving for the three color channels most consistent with the hypothesis-we call these the coefficient color channels. Simulations demonstrate that a von Kries type scaling is an excellent vehicle for color constancy relative to the new coefficient channels. As a second step, we consider experimental evidence for the presence of coefficient channels in the human visual system. A review of the literature reveals many stimulus-response studies that have observed color channels similar to our derived channels.