3D Faces Evoke Stronger fMRI Activation than 2D Faces

Abstract

Despite behavioural evidence that three-dimensional (3D) faces are processed more accurately and quickly than two-dimensional (2D) faces, functional magnetic resonance imaging (fMRI) studies of face processing typically rely on 2D images of faces. Moreover, fMRI studies of 3D vision typically use random dot stereograms, a highly unnatural stimulus. Given the importance of stereopsis in 3D form perception, we expected that neural activation would differ for 3D faces and 2D faces not only in dorsal-stream areas previously implicated in processing depth for simple visual stimuli, but also within face-selective areas in the ventral stream. We used fMRI to investigate brain activation for images of real people presented orthostereoscopically (at the geometrically correct distance and size) with high-quality displays (using a PROPixx MRI 3D projector, viewed through polarized glasses and first-surface mirrors). In the 2D condition, the same image was presented to both eyes, producing zero disparity, as when viewing a 2D picture. In the 3D condition, stereopairs were presented separately to each eye. Stimuli were presented in a block design with a one-back task to maintain attention. Localizers were used to identify face- and depth-preferring regions of interest. Higher activation for 3D than 2D faces was observed not only in depth-selective occipitoparietal cortex (in the caudal intraparietal sulcus) but also in fusiform and occipital face areas, demonstrating that depth information affects processing in both visual streams. Our results suggest that while pictures are a reasonable proxy for studying faces in the real world, models of face processing should consider the impact of 3D form in tasks like face recognition. Moreover, this approach opens new avenues for investigating the contribution of 3D information to category-specific responses in high-level vision.