An image of a face depends not only on its shape, but also on the view
point, illumination conditions, and facial expression. A face recognit
ion system must overcome the changes in face appearance induced by the
se factors. Two related questions were investigated: the capacity of t
he human visual system to generalize the recognition of faces to novel
images, and the level at which this generalization occurs. This probl
em was approached by comparing the identification and generalization c
apacity for upright and inverted faces. For upright faces, remarkably
good generalization to novel conditions was found. For inverted faces,
the generalization to novel views was significantly worse for both ne
w illumination and viewpoint, although the performance on the training
images was similar to that on the upright condition. The results indi
cate that at least some of the processes that support generalization a
cross viewpoint and illumination are neither universal (because subjec
ts did not generalize as easily for inverted faces as for upright ones
) nor strictly object specific (because in upright faces nearly perfec
t generalization was possible from a single view, by itself insufficie
nt for building a complete object-specific model). It is proposed that
generalization in face recognition occurs at an intermediate level th
at is applicable to a class of objects, and that at this level upright
and inverted faces initially constitute distinct object classes.