Face processing is one of the best candidates for modularity, considering the social importance of faces and their physical homogeneity. Neuropsychological, neuroimaging, animal and developmental data all seem to sustain its computational and neural modularity. Nevertheless, this review presents contradictory arguments, suggesting the difficulty of defining a real module for face processing, in terms of localization, specialization and innateness. The most appopriate version, at least for methodological reasons, is that of diffferent gradients of modularity, distinctively acknowledged at the neural and cognitive levels.
KEYWORDS: face processing, modularity, fusiform face area, infant face recognition