Abstract
Studies of color constancy (CC) have frequently used simulated Mondrian patterns to isolate sensory and perceptual processes from long-term memory or expectation (which could affect the perception of natural scenes). Indices of CC in such studies have been wildly variable, spanning the range from 5% to over 80% (as is not untypical of other constancies, e.g. size constancy in Holway & Boring, Amer.J.Psych., 1941, 21–37). We controlled the direction of illumination change and pattern chromaticities to clarify the sources of such variability. Our naïve observers rated how well the central squares in pairs of successive Mondrians matched in material or in hue and saturation. Display duration was 1 sec to limit chromatic adaptation. The simulated Mondrians comprised 49 (7×7) abutting surfaces, whose global illuminants were 4000K or 16000K (first Mondrian) and 6700K (second). Results: a major source of variability is whether the task requires the observer to attend to the proximal stimulus to judge appearance (hue, saturation) by introspection, or to the distal stimulus (the simulated surface property) to form a ‘judgment of origin’ of the real-world change in chromaticity. We speculate that attention selects among chromatic signals which are local to each patch and specify the proximal stimulus, or which encode edge ratios and specify the distal stimulus, these signals probably being available in parallel (Foster et al., 2001, PNAS, 98, 8151).