Abstract
Creating three-dimensional (3D) visual representations of the environment from two-dimensional (2D) retinal images is a complex but fundamental problem the brain solves to support perception and action. Here we present results from electrophysiological experiments conducted in the caudal intraparietal area (CIP) of the macaque monkey to elucidate the role of perspective and binocular disparity cues in the neural representation of 3D object orientation. Using planar surface stimuli, these experiments reveal that individual CIP neurons are selectively tuned for both slant (rotation about an axis perpendicular to the line-of-sight) and tilt (rotation about an axis parallel to the line-of-sight), two parameters defining 3D orientation. Across the population of recorded neurons, the distribution of preferred orientations was statistically indistinguishable from uniform. This property may be beneficial for guiding interactions with objects, consistent with CIP's projection to areas implicated in affordance-based actions such as grasping. Previous theoretical and psychophysical studies show that the relative reliability of perspective (compared to disparity) cues increases with slant. Since accurate 3D perception requires that these cues be integrated according to their reliabilities, the contributions of perspective and disparity cues to the responses of CIP neurons were evaluated. The contribution of each cue was dissociated by measuring planar slant tuning curves using mixed-cue (both perspective and disparity cues, either congruent or conflicting) and cue-isolated (perspective or disparity) stimuli. We find that perspective contributes more to the responses of CIP neurons preferring larger slants, mirroring the slant-dependent reliability of perspective cues. Moreover, some neurons that are sensitive to both cues when tuning is assessed with cue-isolated stimuli are found to disregard one of the cues when they are presented together, but in conflict. These findings suggest that perspective and disparity cues are weighted according to their slant-dependent reliabilities in area CIP to create robust 3D representations of object orientation.
Meeting abstract presented at VSS 2016