Defining face perception areas in the human brain: a large-scale factorial fMRI face localizer analysis

A number of human brain areas showing a larger response to faces than to objects from different categories, or to scrambled faces, have been identified in neuroimaging studies. Depending on the statistical criteria used, the set of areas can be overextended or minimized, both at the local (size of areas) and global (number of areas) levels. Here we analyzed a whole-brain factorial functional localizer obtained in a large sample of right-handed participants (40). Faces (F), objects (O; cars) and their phase-scrambled counterparts (SF, SO) were presented in a block design during a one-back task that was well matched for difficulty across conditions. A conjunction contrast at the group level {(F-SF) and (F-O)} identified six clusters: in the pulvinar, inferior occipital gyrus (so-called OFA), middle fusiform gyrus (so-called FFA), posterior superior temporal sulcus, amygdala, and anterior infero-temporal cortex, which were all strongly right lateralized. While the FFA showed the largest difference between faces and cars, it also showed the least face-selective response, responding more to cars than scrambled cars. Moreover, the FFA's larger response to scrambled faces than scrambled cars suggests that its face-sensitivity is partly due to low-level visual cues. In contrast, the pattern of activation in the OFA points to a higher degree of face-selectivity. A BOLD latency mapping analysis suggests that face-sensitivity emerges first in the right FFA, as compared to all other areas. Individual brain analyses support these observations, but also highlight the large amount of interindividual variability in terms of number, height, extent and localization of the areas responding preferentially to faces in the human ventral occipito-temporal cortex. This observation emphasizes the need to rely on different statistical thresholds across the whole brain and across individuals to define these areas, but also raises some concerns regarding any objective labeling of these areas to make them correspond across individual brains. This large-scale analysis helps understanding the set of face-sensitive areas in the human brain, and encourages in-depth single participant analyses in which the whole set of areas is considered in each individual brain.