Counting Colors: Judgments of Summary Statistics in Color Arrays

Abstract

The visual system can reliably extract the average value of stimulus distributions that vary along many dimensions, from orientation to facial expressions. We examined how sensitive observers were to the average chromaticity of color distributions. This average has special significance for color coding (e.g. to estimate the illuminant) but how it might depend on the level of representation (e.g. perceptual vs. cone-opponent) or nonlinearities (e.g. categorical coding) is unknown. Stimuli were arrays of spots that alternated between two component colors with differing hues that bracketed a given color boundary. Observers set the array so that the composite color equaled a specified reference hue. These settings could be made reliably though variance increased with increasing angular difference between the components. Matches for an average red, blue-red, or yellow-green were consistent with the arithmetic mean chromaticity of the pair, while blue-green settings instead deviated toward blue (indicating greater perceptual weight for the green component). The settings show little evidence for categorical coding of the colors, and moreover cannot be predicted from the scaled red-green and blue-yellow appearances of the individual components, suggesting that there is little influence of linguistic or phenomenal color categories on the mixture judgments.