On “A three-stage color model”

In regard to colored lights, an opponent cell will indeed increase its firing to luminance increments (as well as to equiluminant color increments) but this does not at all imply confounding; rathe, r, in both cases, the cell will signal more color. Alternatively stated, when one increases the luminance (or radiance, or energy, etc.) of a colored light, any theory must predict that its chromaticness (as well as its whiteness) will increase. Otherwise, incrementing the luminance of a colored light would be equivalent to dcsaturating it. For example, within opponent theory, if one increments the luminance of a red light, then not only must nonopponent cells increase firing so there is more whiteness, but also opponent, say R/G, cells must increase firing so that there is more redness, thereby producing at least approximate invariance of the apparent saturation of the light. (Of course, if one increments the light's redness, i.e. purity, at equal luminance, then nonopponent cells will not be affected but opponent cells must again increase firing; to signal increased saturation of redness.) Therefore, for colored lights, the De Valois' statement that an opponent cell will fire to both luminance and color changes is absolutely true, but confounding is not involved--rather, the opponent cell is in both cases signaling more color, as it must.