Abstract
A useful framework for understanding the mental representation of facial identity is face-space, a multi-dimensional cognitive map in which individual faces are coded relative to the average of previously encountered faces, and in which the distance among faces represent their perceived similarity. We examined whether this framework is useful in understanding prosopagnosia, a disorder characterized by an inability to recognize familiar faces, despite normal visual acuity and intellectual abilities. We tested the mental representation of faces in 6 patients with congenital prosopagnosia (CP) and 1 patient with acquired prosopagnosia (AP), and compared their performance to 14 age- and gender-matched control participants. We used digital images of male and female faces and morphed them relative to an average face. To assess whether faces are coded relative to an average face, we examined face identity aftereffects: a shift in perceived identity in the direction opposite to the adapting face, relative to average. To assess the layout of face-space, we measured perceived similarity and distinctiveness of caricatures and anti-caricatures relative to the veridical image. In addition, we used multi-dimensional scaling to analyze the similarity ratings of non-morphed identities. Across 5 behavioral tasks, CP patients revealed a remarkably intact face-space, whereas the AP patient's performance deviated significantly from the control and CP groups. The findings provide distinctions between the behavioral profiles of CP and AP patients, and provide constraints on the utility of the face-space framework in describing how familiar faces are recognized.