Individual faces vary considerably in both the duality and quantity of
the information they contain for recognition and for viewpoint genera
lization. In the present study. we assessed the typicality, recognizab
ility, and viewpoint generalizability of individual faces using data f
rom both human observers and from a computational model of face recogn
ition across viewpoint change. The two-stage computational model incor
porated a viewpoint alignment operation and a recognition-by-interpola
tion operation. An interesting aspect of this particular model is that
the effects of typicality it predicts at the alignment and recognitio
n stages dissociate, such that face typicality is beneficial for the s
uccess of the alignment process, but is adverse for the success of the
recognition process. We applied a factor analysis to the covariance d
ata for the human- and model-derived face measures across the differen
t viewpoints and found two axes that appeared consistently across all
viewpoints. Projection scores for individual faces on these axes (i.e.
the extent to which a face's 'performance profile' matched the patter
n of human- and model-derived scores on that axis), correlated across
viewpoint changes to a much higher degree than did the raw recognizabi
lity scores of the faces. These results suggest that the stimulus info
rmation captured in the model measures may underlie distinct and disso
ciable aspects of the recognizability of individual faces across viewp
oint change. (C) 1998 Elsevier Science Ltd. All rights reserved.