Abstract
Human observers can estimate 3-D object shape using vision or touch. To what extent can visual information about shape be compared to that obtained by touch? Are the representations of 3-D shape derived from vision and touch in a comparable format? J. J. Gibson (1962,1963) described an intriguing cross-modal shape matching task, and concluded that vision and touch “seem to register the same information and to yield the same phenomenal experiences”. Given the obvious sensory differences between vision and touch, finding that they have essentially the same representation of shape would be an important result. It is surprising, however, that Gibson never published any quantitative data that would support this conclusion. Furthermore, he used a set of man-made sculpted objects. The purpose of our experiment was to collect quantitative data using Gibson's cross-modal shape matching task and evaluate his conclusions. An important second purpose was to investigate the validity of his claim using naturally-shaped 3-D objects. On any given trial, observers haptically explored the shape of one of 12 solid replicas of bell peppers (Capsicum annuum) for 3, 5, or 7 sec. The observers then indicated which of 12 visible objects' shape matched that of the haptically-presented object. Each observer participated in 5 sessions (120 trials), enabling an evaluation of the effects of increasing experience with the objects. The results showed that while observers could perform the task at reasonable levels of accuracy (40-75 % correct), significant errors in cross-modal matching still occurred after extensive experience. Some haptically-presented objects were accurately matched to their visible replicas. For many other haptically-presented objects, however, they were frequently confused with visible objects that had different shapes but possessed similar global features. We conclude that the visual and haptic representations of object shape are similar, but different.