Comparison of human and marmoset basic-level face categorization based on shape

Our ability to process faces depends on high-level visual areas in the brain which are thought to be partially shared between human and non-human primates. Nonetheless, studies of face recognition – whether for detecting or discriminating faces – in non-human primates have generally relied on 2D face photographs where low-level strategies relying on texture and appearance cues can be employed to adequately support task performance. Here, we tested humans and marmoset monkeys on a geometry-based invariant face category discrimination task under 3D pose and lighting variation using two textureless faces (human versus monkey) that only differed in their internal face geometry. Both humans and marmosets performed this shape based, basic-level face category discrimination task at comparable levels to their basic-level object category discrimination performance, while state-of-the art artificial vision systems specifically struggled on the face categorization task. Marmosets and humans both exhibited face-specific behavioral signatures including lower category discrimination performance for inverted faces, faces lit from below, and contrast reversed faces compared to normally lit upright faces. However, the magnitude of these deficits differed between marmosets and humans with marmosets less affected by inversion and more affected by contrast reversal than humans. Our results suggest that different primate species may differ in the details of their face processing. However, invariant face categorization ability based purely on shape appears to be generally preserved across simian primates from among the smallest to the most advanced, consistent with the presence of partially overlapping underlying high-level visual brain areas.