Abstract
In a typical color constancy study, the observer adjusts a test surface seen under some illuminant to match a target surface seen under a different illuminant. Do such asymmetric matches predict behavior in a more natural task in which the observer selects an object based on color? We measured constancy using two tasks. In our color selection task, observers viewed a target square against a variegated color background under simulated illuminant D65. The target was surrounded by four test squares under a test illuminant (4500°K or 12000°K): two squares were dissimilar distractors, and two were competitors whose degree of similarity to the target varied. We asked four observers to select the test square closest to the target in color (neutral instructions). For each of four targets, we derived a selection-based match using maximum likelihood difference scaling. We compared these to actual asymmetric matches obtained for the same stimulus configuration. The two types of matches were in good agreement. In both cases, the overall level of constancy was low. We repeated the experiment with three additional groups of observers, each given different instructions (judge the apparent reflectance of the stimuli, the physical reflectance of the stimuli, or the physical spectrum of the light reaching the eye). Performance for the physical spectrum instructions was similar to that for the neutral instructions. The apparent and physical reflectance groups were different. First, observers in these groups showed higher constancy. Second, the asymmetric matches no longer predicted observers’ choices. Third, there was more individual variation within these groups. For our simple stimulus configuration, the relation between asymmetric matches and color-selection depends on instructions. The performance of observers in the reflectance groups suggests that they may employ a different processing strategy from the observers in the other two groups, possibly one involving explicit reasoning.
Meeting abstract presented at VSS 2013