Computing the shapes of object boundaries from fragmentary image contours poses a formidable problem for the visual system. We investigated the extrapolation of contour shape by human vision. Measurements of extrapolation position and orientation were taken at six distances from the point of occlusion, thereby yielding a detailed representation of the extrapolated contours. Analyses of these measurements revealed that: (i) extrapolation curvature increases linearly with the curvature of the inducing contour, although there is individual bias in the slope; (ii) the precision with which an extrapolated contour is represented is roughly constant, in angular terms, with increasing distance from the point of occlusion; (iii) there is a substantial cost of curvature, in that the overall precision of an extrapolated contour decreases systematically with curvature; (iv) the shapes of visually extrapolated contours are characterized by a nonlinear decrease in curvature, asymptoting to zero; and (v) this decaying pattern of curvature is explained by a Bayesian model in which, with increasing distance from the point of occlusion, the prior tendency to minimize curvature gradually dominates the likelihood tendency to minimize variation in curvature.contour completion ͉ curvature ͉ interpolation ͉ occlusion ͉ shape perception A fundamental problem faced by the visual brain in computing object structure is the fragmentary nature of the retinal inputs. Large portions of object boundaries are often missing in the retinal images, either due to partial occlusion or because of insufficient local image contrast. Occlusion in particular poses a ubiquitous problem, given the multiplicity of objects in the world and the loss of one spatial dimension during image projection. To compute object structure from fragmented image data, the visual system must solve two related problems. It must determine (i) whether disparate image elements are in fact part of a single continuous contour (the ''grouping'' problem), and (ii) what shape the contour has in the missing portions (the ''shape'' problem).A great deal of research has addressed the grouping problem in the contexts of partly occluded contours, illusory contours, and discretely sampled contours (1-11). This research has examined the geometric constraints that underlie the grouping of local elements into extended contours, as well as how these constraints relate to the statistics of natural images. By contrast, there has been relatively little psychophysical work on measuring the shapes of the missing portions (12-15). Because the missing portions of contours are synthesized entirely by the visual system, their detailed shapes are likely to be revealing about its underlying constraints and mechanisms.Two constraints have been recognized in computational vision: (i) minimization of total curvature, and (ii) minimization of variation in curvature. Minimizing total curvature ͐ 2 ds (also known as ''bending energy'') tends to make contours as straight as possible and leads to a class of inte...