Abstract
It has been suggested that the relative contribution of shape compared with texture information to face recognition decreases with increasing familiarity. Extending recent findings of a learning advantage for spatial caricatures, we contrasted effects of spatial and texture caricaturing on face learning and recognition. We assessed performance and event-related potential (ERP) effects in a learning paradigm, using different images at learning and test. Based on a set of 94 faces initially digitized using a 3D system (di3D), 3D caricatures were created by exaggerating shape or colour differences of individual faces from gender matched averaged faces (n = 45 and n = 49 for male and female averages, respectively) at a level of 50%. From caricatured and veridical 3D objects we exported six face stimuli in slightly different views. At learning, 36 participants were presented with faces of 45 different individuals. Fifteen identities each were learned using veridicals, shape or texture caricatures. Each identity was presented in three slightly different views. At test, we presented learned faces in the same condition in which they had been learned, but using three different views. An equal number of veridical and caricatured novel faces was presented and participants performed an old/new task. We found performance benefits for both types of caricatures that were further modulated by familiarity: for learned faces benefits were largest for texture caricatures, whereas novel faces profited most from shape caricaturing. Furthermore, at test we found increased occipitotemporal negativity of N170, P200 and N250 components, as well as a more positive late positive component (LPC), for both types of caricatures. Whereas N170 caricature effects were restricted to novel faces, later caricature effects were not modulated by familiarity. Our results imply that distinctive shape and texture facilitate face learning, with distinctive texture contributing in particular to recognition of learned faces.
Meeting abstract presented at VSS 2012