Abstract
A whole network of brain areas showing larger response to faces than other visual stimuli has been identified in the human brain using fMRI (Sergent, 1992; Haxby, 2000). Most studies identify only a subset of this network, by comparing the presentation of face pictures to all kinds of object categories mixed up (e.g., Kanwisher, 1997), or to scrambled faces (e.g., Ishaï, 2005), using different statistical thresholds. Given these differences of approaches, the (sub)cortical face network can be artificially overextended (Downing & Wiggett, 2008), or minimized in different studies, both at the local (size of regions) and global (number of regions) levels. Here we conducted an analysis of a large set of right-handed subjects (40), tested with a new whole-brain localizer to control for both high-level and low-level differences between faces and objects. Pictures of faces, cars and their phase-scrambled counterparts were used in a 2x2 block design. Group-level (random effect) and single subject (ROI) analyses were made. A conjunction of two contrasts (F-SF and F-C) identified 6 regions: FFA, OFA, amygdala, pSTS, AIT and thalamus. All these regions but the amygdala showed clear right lateralization. Interestingly, the FFA showed the least face-selective response among the cortical face network: it presented a significantly larger response to pictures of cars than scrambled cars [t=9.3, much more than amygdala (t=2.6), AIT (t=2.1) and other regions (NS)], and was also sensitive to low-level properties of faces [SF - SO; t=5.1; NS in other areas]. These observations suggest that, contrary to other areas of the network, including the OFA, the FFA is a region that may contain populations of neurons that are specific to faces intermixed with populations responding more generally to object categories. More generally, this study helps understanding the extent and specificity of the network of (sub)cortical areas particularly involved in face processing.