Understanding the shape of 3D objects is particularly significant in scientific visualization and CAD/CAM applications, however objective measurements of how various rendering techniques affect object perception has not received enough attention. We propose a methodology which uses the Vandenberg and Kuse [1978] mental rotation paradigm as a mechanism for evaluating how well viewers are able to encode and match novel shapes presented in a computer graphics display. The established methodology and body of research on mental rotation provides a basis for its use to probe the influence of rendering on 3D shape perception. Our approach is complementary to techniques that evaluate perception of local shape properties and the adjustment of gauge figures to match local surface orientation [Koenderink et al. 1992]. We demonstrate the methodology with an experiment showing that at least in some circumstances, subjects are more accurate at shape perception with Blinn-style rendering than with Lambertian-style rendering.