Abstract
Apparent 3-D depth strongly influences whether the dimensions of a simple shape are perceived separably or holistically. We conducted two experiments in which subjects made judgments about the 2-D planar shape of shape-from-shading images that appeared to have either concave or convex depth depending on the images' orientations. Expt. 1 examined to what degree image parts (as predicted by Hoffman ‘Richards’ minima rule) were perceived independently. Subjects were instructed to identify images based solely on the shape of a single diagnostic part, while the other (non-diagnostic) parts of the image either 1) remained constant across trials or 2) varied randomly from trial to trial. A large difference in performance between these two blocks, an “orthogonal insertion cost”, indicates holistic processing of the image, since parts irrelevant to the task cannot be ignored easily. Subjects showed large orthogonal insertion costs when the images appeared convex and minimal costs when the same images were rotated to appear concave. This suggests that even though the convex and concave stimuli were actually the same image, perception of convexity led subjects to encode the images holistically, while concavity led subjects to encode perceptually independent parts. Expt. 2 was a replication of Expt. 1 with a prosopagnosic patient. This patient, who has documented impairments on tasks requiring holistic encoding of images, showed a distinctly different pattern of performance from healthy subjects. He showed no orthogonal insertion cost for convex images, but a strong cost for concave images. This suggests that in contrast to controls, the patient did not perceive convex shapes holistically, but could and did integrate the parts of the concave images into a global whole. In summary: 1) 3-D curvature influences encoding of a 2-D outline shape, and 2) holistic deficits associated with visual agnosia apply to convex but not concave images.
AC supported by Center for the Neural Basis of Cognition