The rapid and automatic categorization of facial expression changes in highly variable natural images

Abstract

Emotional expressions are quickly and automatically read from human faces under natural viewing conditions. Yet, categorization of facial expressions is typically measured in experimental contexts with homogenous sets of face stimuli. Here we evaluated how the 6 basic facial emotions (Fear, Disgust, Happiness, Anger, Surprise or Sadness) can be rapidly and automatically categorized with faces varying in head orientation, lighting condition, identity, gender, age, ethnic origin and background context. High-density electroencephalography was recorded in 17 participants viewing 50 s sequences with natural variable images of neutral-expression faces alternating at a 6 Hz rate. Every five stimuli (1.2 Hz), variable natural images of one of the six basic expressions were presented. Despite the wide physical variability across images, a significant F/5 = 1.2 Hz response and its harmonics (e.g., 2F/5 = 2.4 Hz, etc.) was observed for all expression changes at the group-level and in every individual participant. Facial categorization responses were found mainly over occipito-temporal sites, with distinct hemispheric lateralization and cortical topographies according to the different expressions. Specifically, a stronger response was found to Sadness categorization, especially over the left hemisphere, as compared to Fear and Happiness, together with a right hemispheric dominance for categorization of Fearful faces. Importantly, these differences were specific to upright faces, ruling out the contribution of low-level visual cues. Overall, these observations point to robust rapid and automatic facial expression categorization processes in the human brain.