Abstract
Faces can be represented as points in a multidimensional face-space with an average face at the center (Valentine, 1991). This average is dynamically updated by experience, resulting in compelling face aftereffects (MacLin & Webster, 2001; Rhodes, et al, 2003; Webster & MacLin, 1999). But does face adaptation have any functional significance? In low-level vision, adaptation can facilitate discrimination around the adapting stimuli (e.g., Clifford, et al, 2001; Phinney, et al, 1997; Regan & Beverley, 1985), although this is not always found (reviewed in Clifford & Rhodes, 2005). Wilson et al (2002) reported reduced discrimination thresholds around the average for synthetic, schematic faces, suggesting a similar functional role for face adaptation. However, subtle discriminations around threshold may not be important for natural face perception. We, therefore, examined suprathreshold discrimination of real faces. Converging evidence from three paradigms showed no enhanced discrimination around the average. Discrimination of interocular spacing differences was not enhanced for faces close to the average. Nor was perceived dissimilarity greater for face pairs close to (spanning) the average. Rather, these were judged to be more similar than other pairs, indicating reduced discriminability around the average. Finally, maximum likelihood difference scaling of similarity ratings (Maloney & Yang, 2003) revealed no enhanced discrimination or “crispening” around the average. Instead, linear functions were obtained, indicating constant perceived differences across the continua. These results suggest that there is no special region of enhanced discrimination around the average face. We consider other possible functions of face adaptation, including neural efficiency and attentional orienting.
NIH EY08266, Australian Research Council