Abstract
Vision science has experienced a recent surge in theories that model perceptual organization as a form of Bayesian inference. Following Helmholtz's principle of unconscious inference, an underlying assumption of such approaches is that perceptual organization is treated as the output of processes that derive the most likely interpretation of the image data. We report a class of illusory contours that seem impossible to explain within such probabilistic frameworks. Thin stationary contours were accreted and deleted by a visible moving surface. Although the image data were fully explained by the visible occluding surfaces in the image, vivid additional illusory contours (ICs) formed on the opposite side of the occluded contours. Matching experiments reveal that the strength of these ICs is a monotonic function of the contrast of the occluding surface and the occluded contours, suggesting that they originate in relatively early stages of cortical processing. In particular, when the contrast of the visible occluding surface increased, the strength of the ICs deceased; a similar decrease in IC strength was observed when the contrast of the occluded contour was reduced. We mapped out the shape of these moving ICs using a dot-probe method, and found evidence that, in opposition to a number of theories of contour interpolation, vivid inflection points and/or corners could be generated by the ICs in these displays. These experiments provide new insights into the mechanisms underlying the formation of ICs in moving images, and challenge Bayesian explanations of contour synthesis in particular, and of perceptual organization more broadly.