Identity-Independent Representations and the Structure of the Expression Space

Abstract

Recent studies using vector flow analysis have shown that expressions of disgust and anger have opposite physical structure to expressions of fear and surprise. Disgust and anger involve compressing modifications of the face surface that, in the case of disgust, serve a sensory gating function. Conversely, fear and surprise involve expanding modifications that, in the case of fear, increase sensory exposure. Consistent with these findings, studies of expression similarity judgements from human observers and computer models have shown that expressions of happiness and sadness also have opposite structure. Taken together, these results suggest the existence of opposing relations between expressions. Is it possible that these relations underlie the neural coding of facial expressions? The present study investigated this possibility. Specifically, I predicted that adaptation to expressions with compressing structure, such as anger and disgust, would bias perception toward expressions with expanding configuration, such as surprise and fear. Experiments 1A and 1B examined cross-identity aftereffects for basic expressions. Following adaptation, participants were shown intermediate morphs between the adaptor and another, structurally opposite expression. Results showed that participants were more likely to categorise these morphs as opposite to the adapting expression. Consistent with previous studies, these effects partly transferred across identity changes, suggesting that adaptation acted on both identity-dependent and identity-independent representations. Because these effects could reflect a generic bias away from the adaptor, Experiment 2 tested the effect of adaptation to anger, disgust, fear, and surprise on perception of six unique morphs representing all possible combinations of the four adaptors. Consistent with my predictions, results showed that anger and disgust equally biased perception toward fear and surprise, and vice versa. To investigate the stimulus selectivity of identity-independent representations, Experiment 3 tested the effect that adaptation to bodily expressions of happiness and sadness had on perception of emotionally ambiguous faces. Results showed that participants were more likely to categorise these faces as sad following adaptation to happiness, and as happy following adaptation to sadness. Taken together, these results strongly suggest that expressions are represented in terms of their structural properties and that emotion adaptation transfers across different stimulus categories.