Abstract
Contours are important to perceive solid shape. Along contours extrema of curvatures (convexities and concavities) specify surface curvature. Vertices of polygons are a special case of extrema assuming that when a vertex is perceived as convex (or concave) its processing is similar to the processing of a positive maximum (or negative minimum). The famous Attneave's cat was described by a set of vertices. There is also a corner enhancement phenomenon: faster responses to probes located near vertices (as opposed to straight contours). We used polygons and their smoothed versions to compare vertices and extrema in two tasks involving global properties of shape. In the smoothed versions a cubic spline removed the vertices. In Experiment 1 observers discriminated stimuli with bilateral symmetry from random stimuli. To test the importance of objectness the contours were either closed to form a single object, or faced each other to form two separate objects. In Experiment 2 observers decided whether a pair of stimuli were identical (translation) or different. In both experiments the presence of vertices or curvature extrema was task irrelevant. Counterintuitively, vertices can make the task harder because the visual system is tuned to processing smooth curvature. Therefore we expected lower performance on reaction time, accuracy and sensitivity (d prime) for polygons. In both Experiments when stimuli were regular (bilateral symmetry in Experiment 1, translation in Experiment 2) smooth contours led to better performance. In our experiments perception of global shape from contours was harder when the convexities and concavities were vertices as opposed to curvature extrema. Note that the involvement of global shape properties in our tasks is critical, as responses to vertices are very fast in a simple detection task. These findings are discussed in relation to theories of shape representation.
Meeting abstract presented at VSS 2015