Abstract
Facial color appearance is an important source of social information. A change in facial skin color can modulate both perceived health and attractiveness, and humans are better at discriminating color of faces than of non-face stimuli. Here we studied how well can humans discriminate changes in color of localized skin features. Four face stimuli in standardized illumination and pose were modified by separately changing the redness (a*), yellowness (b*) and chroma of random clusters of localized spots centered on Laplacian of Gaussian (LoG) maxima. Modified face stimuli were then paired with the original stimulus and observers had to decide which of the two showed a perceivable local change in color. N = 6 observers completed three blocks (a*, b* and chroma changes) of trials regulated by an adaptive staircase procedure. Using a maximum likelihood fit of psychometric curves we characterized the discrimination thresholds for increase or decrease in a*, b* and chroma, respectively. For 75% correct responses the color discrimination threshold was smallest for the increase in redness (Da* = 9.54). Higher were the thresholds for an increase in yellowness (Db* = 13.40) and an increase in greenness (Da* = −14.93), while the threshold for the increase in blueness (Db* < 0) was too high to be reliably measured using our stimulus range with a minimum Db* of −30. Observers were twice more sensitive to an increased chromaticity compared to a decreased chromaticity. Our results suggest that the human visual system is most efficient at detecting increases of chromaticity and in particular increases in local redness, and less efficient at detecting discolorations or changes towards blue and green, both colors that are not associated with skin tone.