Abstract
It is commonly argued that the face-processing system primarily involves configural processing. To evaluate this claim, we designed a task in which subjects are asked to discriminate sequentially presented pairs of faces or houses that vary either in their parts (eyes and mouth in faces; windows and doors in houses) or in the distance among these parts (configuration). Faces and houses were presented either upright or inverted. Importantly, stimuli were created to yield similar performance level for all upright stimuli. Converging evidence from individual differences and from group-based analyses of the behavioral data as well as fMRI data suggest that the face system processes both configural and part-based information of upright faces only and is qualitatively different from non-face mechanisms. Specifically, 1) inversion effect was as great for parts and configuration discrimination of faces, but there was no inversion effect for houses. 2) for upright faces only (not for inverted faces or houses) did we find a correlation across subjects between performance on the configuration and part tasks. 3) fMRI investigation with the same stimuli showed that the FFA responded strongly and equivalently for face parts and configuration, and much less strongly (and equivalently) for house parts and configuration, indicating that this region is stimulus-specific for faces, not process-specific for configuration discrimination. Further, an area responsive to house stimuli showed higher activity for house parts than house configurations. Taking together, our findings suggest that the face-processing system is not a general configuration discrimination system, but a stimulus specific system, with qualitatively different mechanisms from those applied to other objects.