Abstract
Shape information is fundamental to object recognition. Objects can be recognized from shape in the absence of color, relative size, or contextual information. To support object recognition from shape, a large number of distinguishable shape representations must be learned and stored in memory. In 4 experiments presented here, participants learned to recognize 16 novel, 2D contour shapes over the course of 9 alternating training and test sessions. In the first experiment (Exp. 1), half of the subjects learned to recognize closed shapes and half learned to recognize open shapes of equivalent complexity. We tested how the presence or absence of closure affects 2D contour shape learning. Our results show that closed shapes are easier to learn to recognize than open shapes. We then show that the benefit for recognition is due to better encoding of closed shapes (Exp. 2) and not due to easier comparison between closed shapes and their representations in memory (Exp. 3). Finally, we show that potential closure (contours that appear likely to close behind an occluder) does not lead to better recognition performance. In fact, open shapes are more easily recognized than equivalent, occluded shapes (Exp. 4). Together, these experiments suggest that closed shapes are, in general, more easily or more effectively encoded. However, for the benefits of closure to be realized, a shape must be geometrically closed (the visible contour must close), not just perceptually closed (behind an occluder).