Abstract
Color constancy is a robust phenomenon most likely mediated by multiple mechanisms, operating at different levels in the visual system. Hering (1874) suggested that in the natural world, the memory colors of familiar objects may influence the extent to which their colors remain constant under changes in illumination. Here we test this hypothesis using a setup which preserves the natural binocular and monocular cues to 3D shape, while allowing us to adjust the apparent color of real objects. Method. The objects were solid, matt-white-painted styrofoam fruits and vegetables arrayed on a white board (70cm × 50cm) contained within a black box (100cm × 80cm × 60cm) and illuminated by a hidden data projector. Each trial consisted of two phases, separated by a 10-sec blank interval. In the test phase, observers briefly viewed an array containing a test object (either a generic dome or familiar fruit — here, a banana) and a reference disk, under the reference illuminant (D65). In the matching phase, the observers viewed an array of 5 colored flat disks, from which they selected the best match to the remembered test object's color. The illuminant in the matching phase was either D65 (memory task), D40 or D145 (constancy tasks). The test object's color varied between trials. 4 observers took part. Results. In the memory and constancy tasks for the banana, matches for yellow test colors were shifted toward more saturated yellows, while matches to bluish and purplish test colors were shifted towards blue. The matches to identical test colors for the generic dome showed no such shifts, but instead were almost perfectly color-constant. Conclusion. The results are consistent with the notion that the banana's ‘canonical’ yellow memory color interferes both with immediate perception of its real color, and with the constancy of that real color under changing illumination. Color constancy does not depend solely on sensory mechanisms.