Wear of composites can be estimated by the degree of marginal discrepancy between the prepared cavity wall and the occlusal margins of composites. Such evaluations are done on casts by comparing and rating the marginal discrepancy with those on standard casts. We analyzed the reliability of this technique on metal and stone specimens. These specimens contained grooves of different width and depth. For the visual comparison we used stone casts of machined standards of known groove depth. We measured the depths of the metal specimens with a profilometer and made stone casts of these original specimens. Using the stone casts of the standards, five dentists estimated the unknown groove depths on the remaining stone casts. These estimates were done under standardized conditions and repeated by each dentist on five different occasions. The results showed that visual depth evaluations of die stone specimens underestimated the depths when compared with the values measured with a profilometer on the original metal models. One investigator gave significantly different (p less than 0.05) groove depth estimates at different occasions.