Abstract
Measures of color constancy vary widely across different experimental techniques within an individual (i.e., 9–85% in Kraft & Brainard, 1999, PNAS, 96, 307–312) and across different individuals within one technique (i.e., 15–57% in the paper cited above). The first observation shows that a person's apparent capacity to be color constant is highly dependent on the method used to assess that capacity, and that although it might be reasonable to attribute color constancy to an organism as a goal, it might not be reasonable to attribute it to the organism as a well defined, intrinsic characteristic. This dependence on technique is potentially confusing because the term “color constancy” suggests an enduring, all-purpose capacity of the human visual system rather than a transitory phenomenon. Color constancy might be better understood as a property emerging from interactions among subtle forms of chromatic adaptation which could not individually support color constancy in a meaningful way.
The second observation - that measures of color constancy vary significantly across individuals - implies that different individuals adjust perceived color in scenes according to different algorithms, or at least differently tuned versions of the same algorithm. To investigate these differences efficiently, experiments should compare environments in which one aspect of the environment (or potential illuminant cue or signal for chromatic adaptation) is changed, while variation in other aspects is minimized. Recent experiments will be reconsidered within the framework of these two observations.
Supported by EPSRC grant GR/S95558