Abstract
When tested with simple, symmetric, geon-like stimuli, IT cells are more sensitive to differences in nonaccidental properties (NAPs) than metric properties (MPs). However, the images of objects, or object parts, are not necessarily simple or symmetrical. Would this sensitivity to NAPs be manifested with complex, irregular shapes? 119 neurons (2 monkeys) were tested with 4 groups of stimuli: a) simple, symmetric, geon-like stimuli differing in NAPs, b) complex, irregular curved Fourier-descriptor-based stimuli, c) simple, asymmetric, curved Fourier-descriptor-based stimuli, and d) stimuli made by connecting the convexities and the concavities of the latter with straight lines. This resulted in c) and d) differing in a NAP, straight vs. curved, while retaining the same general shape. Stimulus differences between pairs within each group and the pairs composed of stimuli from c) and d) were calibrated. Changing a NAP resulted in a large neuronal modulation which was equivalent within group a) and between groups c) vs. d) and significantly greater than the modulation produced by shape changes within b), c), and d). The equivalence in modulation magnitude within a) and between c) and d) suggests that NAP sensitivity does not depend on symmetry. The low modulation within b), c), and d) suggests that IT cells are more sensitive to NAP than to other shape differences. We extended this study by adding stimulus sets that differed in NAPs but were of progressively greater complexity. This was done by increasing the frequency of the Fourier descriptors in c) and d) above. 31 neurons (2 monkeys) showed a significant reduction in NAP sensitivity at higher complexity levels. The more complex shapes approached texture-like masses in appearance (such as the silhouette of a bush), with short and highly variable contours. Overall, the sensitivity of IT neurons to NAP differences is maintained for moderately complex, asymmetrical shapes but is reduced when the complexity of the shape-outlines renders them texture-like.