Abstract
[Introduction] Shifts in the visual sensitivity introduce shifts in color appearance, which is often experienced under changes in illuminant chromaticity. A few previous studies suggested that the shift in color appearance could be explained by applying a set of fixed coefficient to each cone response, regardless of lightness level. The present study tested whether a set of fixed coefficients is applicable to a wide range of lightness. [Methods] The observer made achromatic settings in a real room with variable-chromaticity illuminant on the ceiling. A matching stimulus was presented on a CRT screen, placed behind a small hole (5cm × 5cm) in a gray-painted wall. The illuminant chromaticity changed from white (D65) to blue, orange, green, and purple. The observer adapted to the illuminant for at least 5 minutes and then started the achromatic settings for five luminance levels of stimulus. Observer repeated the settings for five times in a random order. The results were assessed with relative M-cone weight (Ahn and MacLeod, 1994). [Results] The achromatic point shifted monotonically toward D65, as the stimulus intensity increased. The shifts in achromatic point were fitted nicely with linear regressions on a log-cone-contrast vs. log-luminance-contrast plane. After a careful analysis for the three cone classes, systematic shifts were also found in the slopes and intersections of the fitted lines. [Discussions] The previous reports, including ours (Kuriki, et al., 2000), suggested that a set of fixed coefficient is sufficient to estimate the shift in color appearance under a change in illuminant chromaticity. However, this was an over simplification. After a precise examination for L and M cones, they show slight but consistent change in the degree of nonlinearity (slope in a log-log plot), as well as S cones. This nonlinearity may explain the appearance of Land's two-color projection, which appears trichromatic, even though it is composed of dichromatic lights.