Localized and Distributed Representations of Person Knowledge for Faces

This file contains the supplementary materials only. Abstract: Modular neural models of face processing posit that face-associated person-knowledge is localized in, or accessed through, the ventral anterior temporal lobe (vATL). However, some studies have suggested that person-knowledge is more widely distributed within a larger face-processing network that includes the mid-fusiform gyrus (‘fusiform face area’). Here, we conducted an fMRI study to distinguish whether person-knowledge is localized or distributed by comparing brain responses evoked by synthetic faces, for which participants had learned person-knowledge (biographical facts) and faces for which the subjects had learned only physical facts. After extensive training, participants were cued to recall a particular biographical or physical fact about the upcoming face. In an alternate passive-viewing task, participants were shown the same faces but were not cued to recall biographical or physical facts. Classification analyses (MVPA) were performed on several a priori chosen face-selective regions (ROIs) in the ventral temporal cortex. Within each ROI tested in isolation, MVPA discriminated faces associated with person-knowledge from faces only associated with physical facts. This result is consistent with a distributed model for person-knowledge. However, when tested in a single model to separate shared and unique information, unique information was limited to the left mid-fusiform and vATL posterior (vATL-p) ROIs. Moreover, the feature weights from these two areas showed that only left vATL-p was specialized for processing biographical facts. This latter result was obtained only when these biographical facts were explicitly retrieved in the cueing task. Thus, our results indicate that the left vATL-p represents explicit recall of face-associated person-knowledge