Abstract
An image with boundary contours defined by abutting gratings has precedence in the competition for dominance in binocular rivalry (Ooi & He, 2006). Do illusory contours behave similarly, as well as obey the occlusion constraint in binocular rivalry? To answer this, we manipulated the pattern and shape of the inducing elements (pac-men) of the illusory Kanizsa-square. The left-eye's half-image consisted of two discs on the left side and two aligned pac-men on the right side (forming a vertical illusory contour). The right-eye's half-image consisted of two pac-men on the left side (also forming a vertical illusory contour) and two discs on the right side. The four elements in each half-image spatially corresponded, and were filled with 135deg grating. These elements were placed against a 45deg grating background. (The grating orientation was counterbalanced in the experiment.) When fused, observers experienced a strong predominance ([[gt]]80%) of perceiving a 45deg grating-square in front of four discs (135deg grating), with the perceived depth of the square increasing with the horizontal widths of the pac-men. This percept is remarkable, considering that when viewed in isolation from the context of the Kanizsa display, a 135deg grating disc versus a 45deg grating segment (quadrant of the pac-man) ordinarily induces strong rivalry alternations. Supporting the notion that the strong dominance is caused by the vertical illusory contours (formed by the vertically aligned pac-men), observers experienced strong rivalry alternations when the two pac-men in each half-image were placed diagonally (eliminating each monocular illusory contour). Furthermore, strong rivalry alternations were experienced when the two half-images were exchanged between eyes, and when the pac-men in each half-image were relocated to form horizontal illusory contours. Altogether, our findings indicate that vertical illusory contours can act as occluding boundary contours (to partially occlude the discs) in resolving binocular correspondence and assigning 3D depth.
Supported by a grant from NIH (R01 EY 015804)