Abstract
Although there is agreement that facial expressions are a primary means of social communication amongst people, which facial configurations are successfully visually interpreted within and across cultures is a topic of intense debate that has reached an impasse. This impasse can only be addressed once we know which facial expressions are successfully visually interpreted within and across cultures in the wild, not in controlled lab conditions. Yet, no such studies exist. We present the first large-scale study of the visual perception of facial expressions of emotion in the wild. Specifically, we analyze over 7 million images. We find that of the 16,384 possible facial configurations that people can produce, only 35 are successfully employed to transmit emotive information across cultures, and only 8 within a small number of cultures. Crucially, we find that the visual analysis of these 35 cross-cultural expressions yields consistent visual perception of emotion categories and valence, but not arousal. In contrast, visual analysis of the 8 cultural-specific expressions yields consistent perception of valence and arousal, but not of emotion categories. In addition, we find that the number of facial configurations that are visually interpreted as communicating each emotion category varies significantly. At one extreme, happiness is visually perceived in seventeen facial configurations. At the other end, disgust is only visually identified in a single expression. We also find that the degree of successful visual interpretation of these facial expressions varies significantly. These unexpected results cannot be explained by current models of the perception of facial expressions.
Meeting abstract presented at VSS 2018