The cardinal directions of color space (S, L-M, and LUM) were originally defined by threshold elevations following prolonged contrast adaptation. We introduce a new super-threshold technique to delineate efficiently independent directions in color space. The technique is a variation of a simultaneous contrast demonstration with two physically identical circles (1 deg, 40 cd/m2), one surrounded by a dark annulus (2 deg, 20 cd/m2) and the other by a light annulus (2 deg, 60 cd/m2). In our variation, we sinusoidally modulated the color of both center circles along a line in color space (2 hz). Modulation along the same line in color space as the annular surrounds (in this case, the LUM cardinal axis) produced a compelling illusion that the two circles are modulated out of phase. Modulation along an independent orthogonal axis (in this case, the S or L-M axis) produced no such illusion; the modulated lights appear to be in phase.

We varied the color direction of surrounds from a mid-white center and measured the directions in which the modulating circles are independent of the surround effects (i.e., appear in-phase). The dimensions of the color space were expressed in threshold units measured for each observer. Results: 1) When the color direction of the surrounds was along the LUM axis, observers (n=7) set independent directions within 10 deg of the standard equiluminant line. Observers differed in the standard deviation of the in-phase settings. This inter-observer variability was similar across tasks (method of adjustment and method of constant stimuli) and suggests individual variation may be due to the tuning widths, and not the tuning directions, of the underlying mechanisms. 2) Within the equiluminant plane, independent directions of modulation were always 90 deg from the surround direction with no preference for the cardinal axes. The results will be discussed in terms of multiple higher color mechanisms and adaptive orthogonalization of the color mechanisms.