Abstract
Purpose and Methods: It is widely believed that the visual system tends to divide smooth contours into separate parts at points of minimum curvature (i.e., points of maximum concavity). We investigated the factors determining part-boundaries using a visual decision task, in which subjects were asked to judge whether two marks placed at distant points along the contour were the same or different. It is known that judgments in this task are slower when the two marks lie on different objects, possibly reflecting the increased time required to move attention from one object to the other. We tested whether a similar deficit applies when the two marks lie on the same object but on opposite sides of a putative part boundary. The crucial comparison was between contour segments with identical curvature but differing in the sign of curvature (i.e. convex vs. concave), as only concave curvature extrema are expected to be perceived as part boundaries. Results: We found a significant effect for concavity/convexity, i.e. a deficit for cases where marks were within different parts as compared to within a part. This difference grew with the magnitude of the curvature at the part boundary, corroborating recent proposals that the salience of parts is determined partly by the magnitude of curvature at the concavity. Response latency also generally increased with the magnitude of curvature, independent of its sign. In subsequent experiments, we found that the between-parts deficit disappears when the global configuration is not consistent with a part-boundary interpretation. Conclusions: Part boundaries-as determined by concavity, curvature, and global configuration-create a measurable delay in the movement of attention along a contour.