Abstract
Facial expressions of emotion are widely considered to be universal. However, mounting evidence now shows that gold-standard facial expressio¬n stimuli are not recognized across cultures, suggesting cross-cultural variance in facial expression signals. Yet little is known about how facial expressions vary across cultures due to their complexity as dynamic signals. Here, we address this question using a novel data-driven method and an information-theoretic analysis to precisely identify similarities and differences in facial expressions of emotion. First, we modelled dynamic facial expressions of the six classic emotions – happy, surprise, fear, disgust, anger and sad – in two cultures – Western European and East Asian – using reverse correlation. On each experimental trial, we generated a random facial animation composed of a random sub-set of individual face movements called Action Units (AUs), each with a random movement. Participants (120 total; 60 Western European, 31 females, mean age 22 years; 60 East Asian, 24 females mean age 23 years) categorized 2400 such facial animations according to the six emotions, otherwise selecting ‘other.’ Next, we derived facial expression models of each emotion and for each participant by measuring the statistical relationship between the dynamic AUs on each trial and the participant’s responses using Mutual Information (MI, FWER p < 0.05). Finally, we used MI to precisely identify the AUs that are culture-specific – e.g., in disgust, Nose Wrinkler (AU9) is Western-specific; Lip Stretcher (AU20) is East Asian-specific – and those that are cross-cultural – e.g., smiling (AU12-6) in happy, wide-open eyes (AU5) and mouth (AU27) in surprise. Our results reveal for the first time the specific cultural variances in facial expressions of emotion, thereby advancing knowledge of human facial expression communication and identifying potential sources of cross-cultural communication breakdown.