Abstract
Colors are often strongly associated with different emotional states and facial expressions, and primate trichromacy may be specifically adapted for judging the health and state of conspecifics from the color of their complexion (e.g. Mollon, 1989; Changizi et al., 2006). We examined the influence of color on perceived expression, by examining how a change in color interacted with facial shape cues when subjects classified the expression on face images. Stimuli were frontal-view images of average male or female faces created with Singular Inversions FaceGen Modeller. The faces had a neutral expression or strong expressions of anger, happiness, or fear. Color in the faces was varied by rotating the hue angle and contrast of all pixels relative to the original color or to a grayscale image along cone-opponent or unique-hue axes. In the first experiment, arrays of faces were generated by morphing between each pair of the three expressions. The morph level was varied in a staircase to find the category boundary between expressions. These boundaries were unaffected by either natural or extreme variations in color. In a second task, subjects viewed a neutral-expression face in different colors and chose among 6 basic expressions. This revealed a large effect of color, with redder (+L/−M cone) faces rated more angry and greener (+M/−L cone) faces more afraid. In a final task, we tested for a direct link between color and expression coding by testing for contingent aftereffects when subjects were adapted to alternating red-angry and green-fearful faces. No color aftereffect was visible, consistent with previous reports for color and facial distortions (Yamashita et al., 2005). Together our results suggest that color and shape provide independently coded information about facial expression and that color cues become salient only in the absence of shape cues.