How much does illuminant color affect unattributed colors?

Abstract

Does the light coming from a surface (as opposed to the surface color) appear the same after adaptation to a new illumination as it did before the illumination changed? Many answers have been proposed over the past century, but all have been unsatisfactory. The main measurement problem is to provide a comparison stimulus that is unaffected by the adaptation being tested. My observers used a mental standard. Under 4000-, 6500-, and 10,000-K adaptations (the extremes and the average of daylight) they produced on a CRT unique hues with a constant saturation that was memorized during training. The main evaluation problem is how to determine the theoretical chromaticity shifts that represent illumination invariance for comparison with the data. Like most previous investigators, I used light sources rather than actual surfaces and illuminants. Using a new technique, I determined theoretical surfaces that would have unique hues under the test illuminants. Using Cohen's basis vectors, I derived theoretical reflectances that under 6500 K would produce the chromaticities that the observer chose as unique hues. The chromaticities of those same reflectances under 4000 and 10,000 K are theoretical points representing illuminants-invariant appearance of the light coming from the surfaces. Even for this small range of illuminants the adaptive shifts were too small for invariance, i.e., the appearance of the light was different even after full adaptation. This result sharpens the question of the basis for humans' concept of color as a stable property of surfaces.