Wenderoth and O'Connor (1987b) reported that, although matches to the straight edge of two triangles placed apex to apex revealed an apparent bending in the direction of the chevron formed by the hypotenuse pair (the Bourdon effect), no perceptual unbending of the bent chevron occurred. Using subjective contour figures, Walker and Shank (1988b) found large and approximately equal bending and unbending effects, consistent with two theories that they proposed. In Experiment 1, using adjustable chevron matching and subjective contours, we found that Bourdon effects, equivalent in magnitude to those reported by Walker and Shank, were 4-5 times larger than unbending effects. In Experiment 2, we used a variation of Walker and Shank's measurement technique, in which subjects selected a matching angle from a graded series. We obtained Bourdon effects similar to those in Experiment 1, but much larger unbending effects. Nevertheless, Bourdon effects were significantly larger than unbending effects in one set of data; and in another, Bourdon test means were larger than unbending test means. In both data sets, there was a large and significant pretest bending effect, which enhanced the magnitude of unbending test minus pretest scores. These results were consistent with our theory but not the theories of Walker and Shank. The variance of unbending test matches, 3-4 times that of Bourdon test matches, reflected the task difficulty. We propose that subjective obtuse angle contraction that exceeds real obtuse angle contraction explains the fact that unbending effects are larger in subjective than in real contours.