Local versus global minima in visual pattern completion

The phenomenon of visual completion poses two questions: (1) When will a figure be seen as being in front of another, as opposed to being one of several nonoverlapping figures in the same plane? And (2) exactly what will be seen behind the occluding figure? Buffart, Leeuwenberg, and Restle (1981,1983) recently argued that both questions could be answered by coding theory. Coding theory assumes that aglobal minimum principle is operative in perception, and that perception will result in an interpretation that can be represented by a code with a minimal information load. Contrary to this view, it is claimed here that only those completion and mosaic interpretations that exploit thelocal advantage of continuity of sides at points of intersection or at the common contour will be perceptually salient. When both a locally simple completion interpretation and a locally simple mosaic interpretation can be made, preference will be stronger for the type of interpretation that is globally more simple. To refute Buffart et al.’s claim and to test the hypotheses formulated above, an experiment was performed in which patterns similar to those used by Buffart et al. were presented to subjects, who traced the contours of the figures they saw. Only a very small proportion of the responses given by the subjects were as predicted by coding theory. The data, on the other hand, supported the idea that a global minimum principle operates within the constraints of locally minimal descriptions.