When figure - ground assignment reverses for a given contour, convexities along it become concavi- ties, and vice versa. Baylis and Driver (1995 Journal of Experimental Psychology: Human Perception and Performance 21 1323 – 1342) examined the possible implications of this effect for the perception of symmetry between contours. Using contrast reversal to determine figure - ground assignment, they argued that the ease of detecting symmetry between two contours depends on whether convexities along them match following figural assignment. However, their results might be attributed to the contrast polarity of individual edges instead. We used random-dot stereograms to vary figure - ground assignment for edges while holding contrast polarity constant. By placing the areas on either side of each contour at various depth planes, conditions were created in which the convex parts yielded for related contours either matched or mismatched following figural assignment. Particular depth arrangements which yielded matching convex parts for symmetrical contours yielded mismatching parts for translated contours, and vice versa. The results confirmed that the relative ease of symmetry-detection versus translation-detection depends on the convex parts yielded for each contour following figural assignment. These findings suggest that the perception of symmetry between contours arises only after they have been assigned to a figure on one of their sides, and is based on a higher-level interpretation of each contour in terms of convex parts, rather than lower-level properties of the image.
