The current research investigated depth information contained in the representation of apparently moving 3-D objects. We measured the magnitude of representational momentum (RM) as index of the consistency of object's shape. Experiment 1A found that RM magnitude was higher when shaded convex moving objects shifted to a flat circle than when that shifted to a shaded concave hemisphere. This difference diminished in the case that the moving objects were the concave (Experiment 1B). Experiment 2 confirmed that these results did not stem from luminance information of shading using luminance-polarized circles. Moreover, Experiment 3 showed that RM magnitude was higher when the convex moving objects shifted to blurred convex hemispheres with low-pass filters than when that shifted to the concave hemispheres. The results suggest that internal representations contain incomplete depth information intermediate between those of 2-D and 3-D objects especially for convexity with low-spatial frequency dominance properties in apparent motion.