Abstract
Accurate estimation of object depth is widely believed to underlie the successful execution of grasping movements. Specifically, it has been proposed that retinal disparities play a critical role in grip pre-shaping and stable finger placement on the object surface, under the assumption that stereo information produces veridical depth estimates. If this were the case, then grasp kinematics should be accurately attuned to the veridical 3D structure of objects. Moreover, they should not be affected by the systematic biases that are known to distort perceptual judgements of object depth. Here we tested these predictions by presenting participants with 3D virtual objects defined by various combinations of stereo, motion and texture information at two egocentric distances. In separate sessions, participants were asked to (a) manually estimate the perceived front-to-back extent of the targets or (b) naturally reach-to-grasp the targets. In the perceptual task, we found that stereo-only objects presented at the far distance (45 cm) appeared most shallow; perceptual depth increased (1) by bringing the objects closer to the observer (30 cm) and (2) by adding either motion or texture information to the baseline stereo objects. These findings are consistent with the idea that visual space is compressed and also that depth estimation is non-veridical. Remarkably, the results of the grasping task revealed that the anticipatory opening of the hand followed the same patterns as the manual depth estimates – the grip aperture was smallest when grasping the distant, stereo-only objects and largest when grasping near, multiple-cue objects. These findings show that visual processing of shape for action control shares the same intrinsic biases known to influence depth perception, providing further evidence for common coding of object properties for perception and action.
Meeting abstract presented at VSS 2016