Recognition of Objects Displayed with Incomplete Sets of Discrete Boundary Dots

Most extant theories of shape perception assume or assert that various contour attributes, and in particular, the orientation, curvature and linear extent of the contours provide essential object recognition cues. The present study examined this proposal using discrete dots that marked locations on the outer boundary of namable objects, providing shape-patterns similar to silhouettes. For each shape, the display initially provided only a sampling of the total number of dots in the boundary, and the number of dots was periodically increased until the participant named the object. There were three treatment conditions in which the initial display as well as the periodic increments consisted of continuous arrays (strings) of dots, randomly positioned dots, or evenly spaced dots. Analysis showed objects were recognized with the fewest percentage of dots with the evenly spaced condition, and participants needed the greatest percentage with the contiguous array condition. In many cases objects could be identified when very few evenly spaced dots were shown, thereby providing large spacing between the dots. It seems unlikely that known neural mechanisms could extract contour attributes, e.g., orientation, curvature, and linear extent, from such sparse stimulus patterns, which provides a challenge to the proposition that these are essential shape cues.