Abstract
Studies examining the role of different depth cues in the control of grasping have typically presented three or fewer objects at three or fewer distances. This raises the possibility that participants learn the stimulus set. If so, even poor depth information could appear to support reliable grasping, provided that it was sufficient to identify which of the learned stimuli was presented, leading to an overestimate of the effectiveness of the cue. We examined this for the case of removing binocular information from the visual scene. The stimuli were rectangular objects, on a surface 400 mm below eye height. We measured kinematics of visually open-loop grasps made to stimuli defined by disparity and texture (binocular viewing), or texture alone (monocular viewing). There were two experiment designs: (i) a conventional “blocked” design, in which there were three distances (200, 350 and 500 mm) and three object sizes (30, 45 and 60 mm), and (ii) a “randomised” design in which the distances were randomised (uniform distribution) in the range 200 to 500 mm, and object sizes were randomly selected from the range 30 to 60 mm (in 5mm increments). Consistent with previous findings (see Melmoth and Grant, 2006), peak grip apertures were larger under monocular viewing than under binocular viewing, in both blocked and randomised experiments. However, removing binocular cues had a significantly larger effect in the randomised experiment. This suggests that learned information in the blocked condition improved performance when binocular cues were unavailable. Previous studies may therefore have underestimated the effect on grasping of removing binocular cues. These results demonstrate that using a small stimulus set in grasping studies can produce results that do not accurately reflect the depth information present in the stimulus itself.
B. Keefe supported by ESRC studentship.