On The Relative Effectiveness of Symmetry and Convexity as Figural Cues

Abstract

The effectiveness of convexity as a figural cue varies with context: subjects are increasingly likely to perceive convex regions as figures as the number of alternating convex and concave regions increases from 2 to 8 (57%–89%). Peterson and Salvagio (2008) observed these context effects only when the concave regions were homogenously colored, regardless of the heterogeneity or homogeneity of convex regions. Goldreich and Peterson (2010) modeled convexity context effects with a Bayesian observer that decided whether the displays were 2-D or 3-D assuming that (1) single objects are likely to be homogeneously colored and unlikely to change color only when occluded; (2) multiple objects can be any color; and (3) in 3-D, objects are more likely to be convex than concave. On Assumption 1 homogeneity of color is a background cue. When convex regions were homogeneously colored, however, the prior for 3-D convex objects was stronger than the background cue. We tested whether similar context effects occur with symmetry. In Exp. 1, symmetry's effectiveness as a figural cue increased as the number of alternating black and white symmetric and asymmetric regions increased from 4 to 6 (56%–68%), p = .001, but didn't increase farther with 8 regions (68%). Thus, the bias toward symmetric 3-D objects overcame the background cue when both symmetric and asymmetric regions were homogeneously colored. In Experiment 2, using 8-region displays, we found that the prior for symmetric 3D objects wasn't large enough to overcome the background cue when symmetric regions were homogeneously colored and asymmetric regions were not: Here, subjects saw symmetric regions as figures on only 40% of trials versus 63% of trials when asymmetric regions were homogeneously colored and symmetric regions were not, p < .0001. Thus, the prior is smaller for 3D symmetric objects than for 3D convex objects.