Colored stimuli appear to be brighter than an achromatic stimulus of the same luminance as the colored stimuli. We have studied this effect using similarity judgments in a triadic scaling task. Several instructions and stimulus configurations were used. The color sets contained colors at three luminance levels. When subjects are instructed to attend to brightness differences, the contribution of the brightness axis is strongly increased relative to the contribution of the color channels. Moreover, a correction to the color input is necessary when subjects are instructed to attend to brightness similarities between saturated colors. For a fixed adaptation level, the chromatic input is relatively stronger for the lower luminance levels. This correction is absent in all other investigated conditions. The correction found here confirms an earlier report by Fairchild and Pirotta (1991).