The extent to which faces and written words share neural circuitry in the human brain is actively debated. We provide an original contribution to this debate by comparing face-selective and word-selective responses in a large group of patients (N=37) implanted with intracerebral electrodes in the ventral occipito-temporal cortex (VOTC). Both face-selective (i.e., significantly different responses to faces vs. nonface objects) and word-selective (i.e., significantly different responses to words vs. pseudofonts) neural activity is isolated through frequency-tagging (Jonas et al., 2016; Lochy et al., 2018, respectively). Critically, this approach allows disentangling category-selective neural responses from general visual responses. Overall, we find that 69.26% of significant contacts show either face- or word-selectivity, with the expected right and left hemispheric dominance, respectively (Fig.1A,B). Moreover, the center of mass for word-contacts is more lateral than for face-contacts, with no differences in postero-anterior axis (Fig.2A). Spatial dissociations are also found within core regions of face and word processing, with a medio-lateral dissociation in the fusiform gyrus (FG) and surrounding sulci (FG+sulci;Fig.2B), while a postero-anterior dissociation is found in the inferior occipital gyrus (IOG;Fig.2C). Despite their spatial dissociations in the FG+sulci and IOG, most overlap in category-selective responses is found in these regions (Fig.1C). Critically, in the overlap-contacts, across the whole brain or specifically in the FG+sulci, between-category (word-face) selective-amplitudes showed no-to-weak correlations, despite strong correlations for within-category (face-face, word-word) selective-amplitudes (Fig.3A), and a strong correlation in non-selective general-amplitudes to words-faces. Moreover, substantial overlap and no-to-weak correlations were observed between faces and a control category (houses) known to be functionally dissociated from faces. Overall, we conclude that category-selectivity for faces and words is largely dissociated in the human VOTC, with a limited spatial overlap likely due to the distant recording of dissociated populations of neurons rather than to shared category-selective representations.
Read full abstract