Abstract
It has been suggested that texture information contributes more to familiar face recognition than shape. In two experiments, we tested effects of reduced identity information in either shape or texture on the recognition of personally familiar faces. Stimuli were derived from images taken with a 3D camera system, and both behavioural data and event-related potentials (ERPs) were analysed. In Experiment 1, participants performed a face familiarity task on images of five personally familiar and five unfamiliar faces, respectively. All faces were shown as i) original images (i.e. including both shape and texture information), ii) "shape-only" stimuli (shape "masks" based on the 3D vertices), and iii) "texture-only" stimuli (i.e. flattened surface maps). Performance was best for original faces, followed by "texture-only" stimuli, and worst in the "shape-only" condition. The N250 familiarity effect was largest for original faces and non-significant for "shape-only" stimuli. Experiment 2 used a similar design, with the only difference that "shape-only" stimuli now consisted of the individual face shape combined with an average texture, and "texture-only" faces showed individual texture combined with an average shape. Again, performance was best for original images, followed by "texture-only", and worst for "shape only" faces. In contrast to Experiment 1, significant N250 familiarity effects were found for all three face conditions, but the effect was smallest for "shape-only" stimuli. Furthermore, performance and ERP familiarity effects for all conditions correlated positively with scores in a face learning (Cambridge Face Memory Test, CFMT), and also with a famous face recognition test (Bielefelder Famous Faces Test, BFFT). Overall, our results suggest that recognition of personally familiar faces is mainly, albeit not exclusively, driven by texture information, and that good recognizers are characterized by a larger processing flexibility, possibly enabling them to cope better with reduced identity information in either dimension.
Meeting abstract presented at VSS 2016