Abstract
The "barbell illusion" (Shapiro, 2005) consists of a center bar flanked by a dark rectangle on one end and a light rectangle on the other. As the luminance of the bar modulates in time, a veil of shading appears to travel from one end of the bar to the other. The shading primarily moves in the direction of minimum contrast between the bar and the flanks; i.e., when the bar becomes white, the motion moves towards the light flank. Previous studies have shown that the direction of motion is predominantly controlled by the contrast at the thin (10 min) edge between bar and flank, but can also be influenced by larger-scale contrast; the effect is therefore useful for understanding the interaction of contrast at different spatial scales. Here we blur the contrast edges to understand how the removal of edge information affects the direction of shading. To measure the perceived direction of motion, a dot is placed just below the center bar and oscillates between the two flanks. The phase of the dot is fixed so that the dot is nearest the light flank when the center bar is at maximum luminance. The observers’ task is report whether the shading motion is in or out of phase with the dot. When the bar is front of the flanks the motion is always towards minimum contrast, when the bar is in a layer behind the flank, motion is primarily perceived in the opposite direction. Haploscopic investigations (modulating bar in one eye, flanks in the other) find a reversal of direction on a substantial portion of trials. A motion energy model suggests the possibility of shading motion being determined at multiple spatial scales and it is possible that the strength motion signals may, in part, be determined by object depth plane.
Meeting abstract presented at VSS 2013