Comparing face viewpoint, expression and identity selectivity in fMRI-defined face patches of macaque frontal cortex

Perceiving and interpreting facial information such as identity, expression, and head orientation is essential for primates as these features provide important social communication cues. To gain insights into the underlying neural mechanisms processing these facial features in prefrontal cortex, we conducted single- and multi-unit recordings in 4 fMRI-defined face patches in 3 macaques: POa and POp in orbitofrontal cortex, PA in ventrolateral prefrontal cortex, in addition to face patch AM in anterior inferotemporal cortex. In each face patch, we found face-selective neurons tuned to identity, expression, and head orientation. A large fraction of these neurons was sensitive to head orientation irrespective of identity or expression. While face-selective neurons preferentially tuned for expressions were mostly present in POa, most face-selective neurons tuned for identity resided in AM. Surprisingly, not only face-selective but also non-face-selective neurons carried similar information about these facial features. Most neurons within each face patch exhibited visual response latencies that were comparable for different face features. At population level, visual response latencies for faces were similar (~70 ms) for both orbitofrontal (POa and POp) and anterior IT (AM) face patches. In prefrontal face patch PA, however, most cells responded much faster to faces, as quick as 30 ms. Neurons generally exhibited fastest face-selective responses (face contrasted with objects), followed by selectivity to head orientations, and yet later for different expressions and identities. While the range of visual response latencies in each face patch is relatively small (interquartile range (iqr): 30 - 80 ms), the selective responses to various face features showed a considerable variation (iqr: 80 - 190 ms). These findings reveal complex prefrontal face-processing signals potentially involving multiple and parallel feedback loops with different areas, prompting a reconsideration of the role of the face-processing system in representing face viewpoint, expression and identity.