Abstract
When an observer moves through the world, stationary objects in a scene have retinal image motion that depends on distance. Theoretical work[1] has shown that the depth of an object in a scene can be computed monocularly from the ratio of retinal image motion and the rate of change of eye orientation relative to the scene. The latter quantity can come from efference copy signals about the smooth pursuit velocity of the eye, and we have previously shown that neurons in area MT can combine ambiguous retinal image motion with smooth pursuit signals to represent depth[2,3]. However, it is also possible that changes in eye orientation relative to the scene can be signaled through purely visual mechanisms. As the eye changes orientation relative to a scene (e.g., during pursuit), dynamic changes in perspective introduce a component of "rocking" motion in the optic flow field (hereafter 'dynamic perspective'). We hypothesized that neurons in area MT combine dynamic perspective information with ambiguous retinal image motion to represent depth in the absence of pursuit or binocular cues. Fixating animals viewed a depth-sign ambiguous motion parallax display with or without a background that contained dynamic perspective. More than half of MT neurons show significant depth-sign selectivity driven by dynamic perspective cues. Moreover, the depth-sign preference is well correlated with that obtained when animals actively pursue a visual target without background motion. These results suggest that the visual system can use dynamic perspective cues as a proxy for efference copy regarding eye velocity.
Meeting abstract presented at OSA Fall Vision 2012