Abstract
The processing of emotional facial expressions has been studied mainly with stereotypical face stimuli, but contextual information leads to drastic modulation in the categorization of facial expressions (Aviezer et al., 2017). In the human brain, brief facial expression changes are quickly read from faces (Dzhelyova et al., 2017). Yet, how reliably these changes are detected with realistic faces embedded in a natural context remain unknown. In this study, faces varied in viewpoint, identity, gender, age, ethnic origin and background context. We recorded 128-channel EEG in 17 participants while they viewed 50s sequences with a neutral-expression face at a rate of 5.99 Hz (F) at two faces orientations (upright, inverted). Every five faces, the faces changed expression to one of the six basic expression (fear, disgust, happiness, anger, surprised or sadness; Ekman, 1993), one emotion per sequence (e.g ANeutralANeutralANeutralANeutralANeutralBExpressiveANeutral …). EEG responses at 5.99 Hz reflect general visual processing, while the EEG responses at F/5 = 1.1998 Hz and its harmonics (e.g., 2F/5 = 2.3996, etc.) index detection of a brief change of natural facial expression. Despite the wide variety across images, a F/5 response was observed for each individual participant, pointing to robust facial expression categorization processes. At the group-level, the categorization response was measured over occipito-temporal sites and was largely reduced when faces were inverted, indicating that it reflects high-level processes. Despite evidence (Leleu et al., 2018; Hinojosa et al., 2015) suggesting that sad expressions are more subtle and thus lead to weaker responses than other emotions, our observations with natural expressions highlight a stronger response for this emotion, especially over the left hemisphere. Moreover, we observed a right hemisphere dominance for a shift from neutral to fearful faces and a left hemisphere dominance for a shift from neutral to happy faces.