Second, we have studied how color a

Second, we have studied how color appearance depends on spatial pattern. The color appearance of a spatial squarewave pattern depends on the spatial frequency of the squarewave. We have found that the change in color appearance with spatial frequency can be explained by assuming that signals from three opponent-colors mechanisms are scaled by a gain factor that depends on the local spatial frequency content of the image. This model is consistent with common engineering practice, in which it is well known that color appearance becomes progressively desaturated as the spatial frequency of the stimulus increases. We have captured this qualitative observation in laboratory measurements and modeled the phenomenon quantitatively (Poirson and Wandell (1993) and Poirson and Wandell (1996)).

Here, we report on new measurements of how color appearance depends on spatial pattern. Our initial measurements were based on simple squarewave stimuli. A general theory that explains the color appearance of more complex texture patterns must be able to predict the appearance of patterns formed by the sum of squarewaves from measurements of the appearance of the squarewaves in isolation. So, in this study, we made measurements of the appearance of squarewave gratings and their spatial mixtures. The squarewaves and their sums were presented on a uniform background field. They were shown in a common phase and orientation, but they were varied in color and contrast.

To measure the color appearance of our test patterns, subjects adjusted the appearance of a uniform matching box to match the appearance of different bars within the test patterns. We tested two empirical properties of these asymmetric color-matches. First, we tested whether the matches satisfied contrast-homogeneity. Suppose that a bar within a pattern formed by the mixture of squarewaves is matched by a uniform matching box. If we double the contrast of the test pattern, will the contrast of the matching box also double? We confirm Poirson and Wandell's (1993) observation that contrast-homogeneity holds for simple squarewaves; we extend their observations to show that homogeneity also holds for sums of low frequency squarewaves. Second, we tested whether the matches satisfied pattern-superposition. Suppose that one of the two bars comprising a squarewave is matched by a uniform matching box, and that a bar in a second squarewave is matched by a second uniform matching box. Now, form the sum of the squarewave gratings such that the two bars superimpose. Will subjects match the superposition of the bars by the superposition of the matching boxes? Again, for the low spatial frequency range we examined, we find that this superposition test holds reasonably well. These results indicate that, to a first approximation, the matches can be described using linear models.