Face perception, perhaps the most highly developed visual skill in humans,
is mediated by a distributed neural system in humans that is comprised of m
ultiple, bilateral regions. We propose a model for the organization of this
system that emphasizes a distinction between the representation of invaria
nt and changeable aspects of faces. The representation of invariant aspects
of faces underlies the recognition of individuals, whereas the representat
ion of changeable aspects of faces, such as eye gaze, expression, and lip m
ovement, underlies the perception of information that facilitates social co
mmunication. The model is also hierarchical insofar as it is divided into a
core system extended system. The core system is comprised of occipitotempo
ral regions in visual cortex that mediate the visual analysis of faces. In
the core system, the representation of invariant aspects is mediated more b
y the face-responsive region fusiform gyrus, whereas the representation of
changeable aspects is mediated more by the face-responsive region in the su
perior temporal sulcus. The extended system is comprised of regions from ne
ural systems for other cognitive functions that, can be recruited to act in
concert with the regions in the core system to extract meaning from faces.