Abstract
Major theories of human visual processing postulate the existence of independent visual systems for action and perception, notably rejecting the more parsimonious assumption that common visual encodings underlie both functions. To investigate the dual-streams hypothesis, previous studies have tested whether grasping movements directed at objects in visual illusions are affected to the same extent as perceptual reports of object size. In many studies, grasp control has been found to resist illusions, apparently supporting the idea that vision for action separately computes accurate size representations. However, an alternative explanation is that repetitive grasping of a small set of target objects (which is ubiquitous in such studies) induces sensorimotor adaptation that washes out the illusion effect on grasping despite the distorted visual encoding. In support of this novel account, we show that repeated grasping of objects in a size-distorting illusion produced increasingly accurate grip pre-shaping across 20 trials while the illusion's effect on conscious perception persisted. Examining this learning process on a trial-by-trial basis, we found that grasping errors induced by a switch in the illusory context were immediately corrected when the next grasp was directed at the same object. Critically, this one-shot learning shows that the gradual reduction of the illusion effect over multiple trials does not involve iterative updating of a separate object-size representation, but rather that it may emerge from a memoryless error correction process. In line with this proposal, conscious perceptual estimates of object size obtained prior to each grasp were found to be the best predictor of the maximum grip aperture (MGA), indicating common visual encoding of object size for perception and action. Our findings demonstrate that when feedforward visual signals convey inaccurate size information, visuomotor networks will integrate visual and haptic prediction errors to achieve appropriate movement calibration.
Meeting abstract presented at VSS 2016