Abstract
In previous research (VSS 2007) we reported that perception of depth and figure-ground organization is strongly affected by Gradient Cuts (GCs): edges that intersect (“cut”) the equiluminance contours of a shading gradient. In particular, GCs biased the gradient side to appear as a farther ground when the size of the angle between the EquiLuminant Contours and the shared edge (the ELC angle), is greater than zero, with the probability of the GC side being seen as closer/figural dropping essentially to zero when the ELC angle reaches about 15°. In Experiment 1 of the present research we show that the ELC angle cannot account for substantial variations in figural/depth judgments that occurred when we manipulated the alignment between the concavities/convexities of a triangular zig-zag edge and the dark and light striations of a multi-cycle gradient region that was cut by the edge at angles substantially greater than 15°. Using bipartite images with central shared edges defined by symmetric and asymmetric triangle waves, we found that alignment of GC edge concavities with the dark striations produced a strong figural bias, and that the alignment of GC edge convexities with the dark striations produced an opposite ground bias. Experiment 2 ruled out an alternative explanation in terms of striation width, since the dark striations in Experiment 1 were thinner than the light striations. Gratings with equally wide light and dark striations and gratings that were contrast-reversed versions of those in Experiment 1 showed that striation polarity (light vs. dark) dominated striation width (wide vs. narrow) in governing these figural/depth effects. We model the data in terms of the number of junctions of various types (light or dark T-, Y-, and arrow-junctions). We also provide an ecological argument about why alignment of edge-concavities and convexities with light and dark striations should matter.