Abstract
The hue of most wavelengths changes when mixed with white light (the Abney effect). We show that these hue changes are qualitatively predicted if the visual system assumes that the desaturation results from a change in spectral bandwidth rather than a spectral dilution. Because of the filtering imposed by the receptor spectral sensitivities, broadening the bandwidth around a fixed peak wavelength alters the ratio of cone responses, and thus the peak must shift with increasing bandwidth to maintain a constant cone ratio. However, we found that observers instead choose the same peak when matching the hue of narrow and broad stimuli with roughly Gaussian spectra, consistent with a nonlinear adjustment that compensates for the eye's filtering effects. When the same chromaticity is instead formed by mixing a wavelength with white, this compensation results in an erroneous hue shift because the “right” response is applied to the “wrong” stimulus. This predicts curved or straight hue loci depending on both wavelength and individual differences in spectral sensitivity, and these are qualitatively consistent with measures of the Abney effect. Our model suggests a novel functional account of Abney effects and color appearance: in which hue is tied to a fixed expected property of the physical stimulus (eg spectral centroid) rather than a fixed mechanistic property of the observer (eg the cone ratios). Such effects also support recent suggestions that natural spectra may in some cases be better described by Gaussian than linear models (MacLeod and Golz, in Mausfeld and Heyer 2003).