Abstract

Krauskopf and his colleagues have proposed a color space with three cardinal axes—the differences in activation in long-and medium-wavelength selective photopigments (an R-G cone axis), the activation in short-wavelength selective photopigments (a B-cone axis), and the combined activation of all three photopigments (an achromatic or R + G + B axis)—each of which is based on a linear combination of cone excitation. We have studied performance in several different visual tasks as a function of chromatic axis. Results from tilt aftereffects, size aftereffects, an opposing motion task, and a color shift task support the hypothesis of independent coding of visual information in at least three mechanisms that correspond fairly closely to the suggested cardinal axes. There is, however, a basic problem in the geometry of this color space. Specifically, although the two axes of the equiluminant plane (R-G and B) are mutually orthogonal, the achromatic axis is not, in fact, orthogonal to either of the two color axes. It is the luminance direction (the R + G direction) that is orthogonal to the equiluminant plane and the luminance direction is not an achromatic direction: R + G varies along the luminance direction but B does not and therefore hue must change. Even though these axes are not orthogonal in the color space suggested by Krauskopf, the performance in the tasks we have examined does show functional orthogonality. Tests of response linearity for each axis may point to possible color spaces that do have orthogonal axes.

© 1989 Optical Society of America