Abstract
Symmetry has long been considered a configural cue, but we failed to find evidence for this claim in previous experiments using brief exposures, leading us to suppose that other processes were necessary to bootstrap the symmetry cue. In our displays, symmetric regions with rectilinear stepped contours alternated with same-area asymmetric regions with the same contour type. Similar results were obtained with curved contour displays. Here we examine whether our results reflect a failure to correctly group the bounding contours of the symmetric regions. In Exp. 1 we tested whether correct contour grouping is necessary for symmetry to operate as a configural cue by manipulating the shape of an asymmetric region sharing one rectilinear stepped contour with a symmetric region. The asymmetric region's non-shared contour either was also a rectilinear stepped contour or it was a curvilinear contour comprised of local curves that were all either convex or concave. We presented such displays for 100 ms with a probe square located on either the symmetric or the asymmetric region. Subjects reported whether the probe was located “on” or “off” the region they saw as the figure at the border shared by the two regions. The symmetric region was more likely to be seen as figure when the unshared edge of the asymmetric region was curvilinear (64% and 68% for concave and convex curvature), than when it was rectilinear (51%). In Exp. 2 we investigated whether these effects were due to grouping alone or to symmetry by replacing the symmetric regions in the Exp. 1 displays with asymmetric regions. Subjects were no more likely to see the asymmetric region as figure when the non-shared contour of the asymmetric region was curvilinear (49%) than when it was rectilinear (51%). Thus, when the two contours of the symmetric region were similar and different from the third contour in the display, symmetric regions were more likely to be seen as figure, implicating grouping as an essential factor.