Abstract
Abstract: :
Purpose: In the ecology of occlusion, changes of egocentric distance affect the absolute size of gaps between T-junctions, while the support ratio of interpolated contours remains constant. In such conditions scale invariance of visually interpolated shapes would be a property of an ideal interpolation process. Gerbino & Fantoni (ARVO 2000) showed that human observers’ estimations of occluded vertices are not scale invariant. Methods: Here, we further demonstrate that interpolated contours are scale dependent using a kinetic pattern containing a diamond whose vertices are partially occluded by disks. When the pattern undergoes a rigid size transformation, all observers reports that the diamond appears more and more rounded as the retinal size decreases. According to our field model (Fantoni & Gerbino, VSS2001) the Minimal Path (MP) strength increases as the retinal gap decreases, while the Good Continuation (GC) strength is simply scaled with size. To measure how retinal size affects the relative strength of MP vs GC we run an experiment using displays in which a 90 deg vertex was occluded either symmetrically (MP-GC angle= 45 deg) or asymmetrically (MP-GC angle= 35 and 25 deg). The factorial design included two levels of support ratio (71 vs 82%) and three observer distances (58, 116, 174 cm), corresponding to retinal gaps of 3, 1.5, and 0.8 deg respectively. To estimate curvature of the interpolated path, observers set three straight line probes at different orientations, one parallel to the MP segment and two parallel to the bisectors of GC-MP angles. Observers were required to set probes tangent to the amodal contour. Results: Data were consistent with predictions of the field model in all retinal gap conditions. Interpolations of small retinal gaps (1.5 and 0.8 deg) were significantly flatter than the interpolation of the large retinal gap (3 deg). Support ratio affected the interpolation path only when the pattern was small, according to the hypothesis that long-range effects are obtained only if elements fall within the field of visual attention. In such cases, the interpolation path becomes flatter when the support ratio is smaller. Conclusion: The field model has been modified to account for the fact that the retinal separation affects the relative strength of MP vs GC only when the pattern is large. The definition of support ratio should be modified to include a reference to the attentional field size. Further work should clarify how interpolation trajectories are affected by the size and distribution of visual attention.
Keywords: 579 shape, form, contour, object perception • 578 shape and contour • 510 perceptual organization