Abstract
The dedicated face-selective patches positioned in the macaque visual cortex are assumed to play a pivotal role in social cognition by processing facial cues during social interactions. However, the extent to which activity in these face-selective patches is driven by more naturalistic, social interactions has been understudied. Here we used contrast-agent enhanced, high-field (4.7T) functional MRI to measure movie-evoked activity in the visual cortex of three adult rhesus macaques (Macaca mulatta). While in the scanner, these animals were shown movies of different social interactions that varied in valence (positive vs. negative), arousal (high vs. low) as well as social direction (interaction vs. observation). While motion sensitive areas such as MT were activated more by movies classified as high arousal (e.g., aggressive fights between conspecifics and mounting behaviors) than movies classified as low arousal (e.g., grooming behaviours), the face-patches, defined using independent localizer data, were activated more by movies classified as interactions (e.g., a conspecific directing affective behavior towards the subject) than movies classified as observations (e.g., multiple conspecifics directing their behavior towards each other, and not the subject). Multi-voxel pattern analysis employed at the level of the inferior temporal cortex, defined as the combination of the cytoarchitectonic regions known as TE, TEO, and the Superior Temporal Sulcus, showed that interaction movies evoked a different pattern of activity across voxels than observation movies. We used a functional brain connectivity analysis, seated on the canonical face-patches, to decouple the neural circuitry processing direct social interactions from the neural circuitry processing third-party interactions. Taken together, these results reveal evidence that the visual cortex has two modes of operation, both supporting social cognition under different circumstances; interaction mode is triggered by the subject’s direct inclusion in a social interaction, whereas observation mode is triggered by watching third parties interact with each other.