Abstract
Despite the fact that we know a great deal about the processing of binocular disparity for synthetic stimuli (such as random dot stereograms) in early stages of visual processing (e.g., V1, V2), little is understood about the contribution of stereopsis to the processing of more naturalistic stimuli in later stages of the visual hierarchy. Here we used functional magnetic resonance imaging (fMRI) to investigate how binocular vision contributes to high-level neural processing of visual objects. Participants viewed objects and scrambled objects (along with faces, scenes and bodies, not discussed here) as 2D or stereoscopic 3D images with a high-quality 3D MRI projector and polarized eyewear. High-resolution 3D objects models were used and displayed using the average human interpupillary distance. The objects were re-scaled and rotated to attain a similar disparity range across the set. Blender software was used to segment the objects into similar parts; importantly each fragment preserved its original disparity profile. The parts were then redistributed to create a cloud of fragments. In the 2D condition one of the images of the stereopair was presented to both eyes. Additionally, we ensured that the viewing geometry was consistent with natural viewing of real objects. Our results show stronger activation for 3D stimuli versus 2D stimuli in both the dorsal and ventral visual streams. More surprisingly, 3D viewing decreased object-selectivity (objects – scrambled objects) in shape-selective regions such as LOC and V3A. This modulation in selectivity was due to a greater increase in activation with the addition of stereopsis (3D – 2D) for scrambled than intact objects. Our results suggest the high-level regions that process different visual categories might be differentially sensitive to availability of binocular disparity.