Abstract
Nonrigid motions frequently occur in nature. According to Koenderink and van Doorn (1986), bendings are “the most general class of deforming motions that permits a solution of the problem of shape from motion”. Despite this fact, bending remains essentially uninvestigated. In our experiment, on any given trial observers viewed two bending rods (defined by the motions of 50 luminous points), and were required to judge which rod was more elastic (i.e., less “stiff”). The stimulus displays were created according to the methods described by Craft, Payne, and Lappin (1986). The method of constant stimuli was used to obtain elasticity difference thresholds. Thresholds were obtained for six experimental conditions formed by the combination of two standard elasticities and three bending planes. The rods bent in either the fronto-parallel plane or bent in a plane that was oriented 42.5 or 85 degrees from fronto-parallel (thus the rods in those two latter conditions bent in depth towards and away from the observers). All observers reliably perceived the bending motions, even those that bent in depth — thus, the perceptual mechanisms that extract 3-D structure do not require rigidity. The observers' average elasticity difference threshold (i.e., Weber Fraction) was 7.6 percent of the standard. The observers' thresholds for bending in depth were quantitatively similar to those obtained for bending in a 2-D image (i.e., the fronto-parallel plane). The overall results reveal that human observers exhibit a moderately high sensitivity for detecting deviations in the elasticity of bending objects.