Concurrent face-selective neural activity across the ventral occipitotemporal cortex revealed with human intracerebral recordings

Abstract

Recognition of object categories from the visual environment is a fundamental brain function, supported in humans by a bilateral network of regions extending ventrally from the occipital pole to the anterior temporal cortex. While dominant models of visual recognition describe this network as being organized hierarchically, with distinct processing stages increasing in representational complexity, abstraction and timing from posterior to anterior regions, direct evidence for this hierarchical organization in the human brain is lacking. Here we take advantage of the high temporal and spatial resolution provided by intracerebral recordings in a large sample of participants (N=140) to provide an extensive characterization of the time-course of category-selective responses to natural images of faces – arguably the most familiar and ecologically valid stimulus in the human environment – across the whole ventral occipitotemporal cortex (VOTC). Objectively tagged face-selective responses in the high-frequency broadband range (30-160 Hz) were found in 10% of recording sites distributed over the whole VOTC, with regional peaks of activity around and along the length of the fusiform gyrus, leading to three key observations. First, face-selective neural activity is characterized both by signal increase (57% of sites; 65% of total response amplitude) and decrease (43% of sites; 35% of amplitude) relative to other object categories, with a respective lateral-medial distinction along the VOTC. Second, an increase in degree of selectivity is observed along the postero-anterior axis, with about 50% of exclusive response to faces in the anterior regions of the ventral temporal lobe. Third and most importantly, face-selective neural activity occurs largely concurrently (80-100ms onset – 350-450ms offset) across the VOTC, with highly overlapping time-courses between regions. Overall, these observations directly challenge the standard view of a hierarchical neural organization of visual object recognition in the human brain and call upon alternative neurofunctional models of this key brain function.