From structure to concepts: The two stages of facial expression recognition

Facial expressions are the dominant way of human emotional and social communications. However, it remains unclear that how can perceivers extract emotion from the face. In the present study, we adopted a repetition-priming paradigm in combine with event-related potentials (ERP) to examine neurocognitive processing stages of facial expression perception. Results showed that emotional words were recognized faster than emotional faces, when both of which were primed by emotional faces, indicating the involvement of concepts processes in facial expression recognition. ERP results showed that all emotional faces could evoke responses of N170, while there was no significant difference between positive and negative emotional faces, suggesting that geometrical configurations of faces rather than emotional concepts of faces are processed at the stage of N170. In contrast, both emotional words and faces showed larger P2 in response to anger than happiness, which suggests that emotional concepts are extracted from faces at the stage of P2. To examine the underlying dynamic causal connectivity between facial structure and emotional conception, we conducted information flow analysis, which showed significant decreases of information flow from the fusiform gyrus to dorsal anterior cingulate cortex/dorsal medial prefrontal cortex and increases of information flow from the fusiform gyrus to posterior insula. These results revealed neural mechanisms underlying processes from physical structure to emotional concepts. Our findings suggest that facial expression recognition consists of two stages from geometrical structure of faces to emotional concepts of facial expressions, which provides evidence for facial expression processing and has important implications in the diagnosis and treatment of emotion recognition related disorders. •Emotional words were recognized faster than emotional faces primed by emotional faces.•Angry emotional words evoked larger N170 than positive ones, while not for emotional faces.•Both emotional words and faces showed larger P2 in response to angry than happy.•The information flow of facial structure transformed to emotional concepts.•Facial expression recognition consists of two stages from facial structure to emotional concepts.