Abstract
Opponent process theory assumes that all colors can be described as different mixtures of the underlying opponent primaries red vs. green or blue vs. yellow. Hue scaling is widely used as a method for estimating the underlying opponent response functions by asking observers to explicitly decompose each hue into the percentage contributed by the different primaries. However, this decomposition can be nonintuitive and there are inherent problems in estimating and analyzing proportions. We developed an alternative hue scaling task based on a compass with the text labels “R”, “B”, “G”, and “Y” placed along the cardinal axes. Observers judged each hue by setting the compass needle to reflect the perceived distance from these landmarks. The hue percentages could then be estimated from these similarity judgments. We tested 30 students on both the compass angle and primary percentage task, repeating each task twice in a counterbalanced order. The stimuli consisted of 36 chromatic angles at 10-deg intervals and constant contrast in a scaled MacLeod-Boynton space. Settings were analyzed in terms of both the mean responses and the variability. Our results indicate that the two techniques are similar with comparable variance, with between-subject differences larger than within-subject. The estimated chromatic response functions were also similar for the two tasks. Both showed weak categorical biases where there were tendencies to give greater weight to the perceptually closer primary or axis. Factor analyses of the variations across observers also showed for both tasks that these differences reflected multiple factors narrowly tuned for different hue angles, and thus both were inconsistent with variations arising from the putative underlying opponent dimensions. The similarities between the tasks suggest that the resulting chromatic response functions are not strongly shaped by the specific nature of the task and thus may reflect the properties of the perceptual encoding of color.