Abstract
Pizlo and colleagues have proposed an interesting new theory of 3D shape perception, in which the bilateral symmetry of objects is exploited to provide veridical estimates of 3D structure. They have also made a surprisingly bold assertion that accurate shape perception is only possible for objects that are symmetrical. The present study was designed to address two important questions related to these claims: First, are observers' judgments of symmetrical objects truly veridical, or are they subject to the same types of distortions that that have been reported previously for the visual perception of 3D shape (e.g. Todd & Norman, 2003); and second, are shape judgments of symmetrical objects significantly more accurate than those obtained for asymmetrical objects? The stimuli consisted of randomly generated, smoothly curved, elongated objects whose orientations in depth were varied across trials.These objects could be either symmetrical or asymmetrical, and they were depicted with a mesh of contours that were arranged both radially and perpendicular to the axis of elongation. The observer's task on each trial was to adjust the orientation of a line to indicate the apparent slant of the object in depth. The results showed clearly that observers make systematic errors in judgements of object slant. Moreover, there were no detectable differences between the judgments obtained for symmetrical and asymmetrical objects.These findings indicate that several of the claims made by Pizlo and colleagues are greatly overstated. Although there may be a strong perceptual bias to prefer symmetrical over asymmetrical interpretations, we found no evidence to suggest that judgments of 3D shape are in anyway dependent or improved by a symmetry constraint.
Supported by NSF BCS-0962119.