The evaluation of margins of restorations in clinical trials relies on the subjective assessment of evaluators, with no instrument having been developed, let alone validated, to assist in the process. The purpose of the present study was to assess the quality of evaluations of marginal adaptation by analyzing the distribution of marginal steps rated according to clinical criteria. Replicas of 435 restorations, the marginal qualities of which had been evaluated according to modified USPHS criteria, were randomly selected from the Occlusin multi-center clinical trial program. The marginal step height in the most deteriorated area of each restoration was measured by means of a digital step-height instrument. An overlap between the steps in the restorations with A (Alfa; replacement unnecessary) and B (Bravo; replacement questionable) ratings was found to be in the range of 101 microm to 321 microm. When the steps with a height greater or less than one standard deviation of the mean were excluded, the overlap was reduced to a range of 168 micro m to 173 microm. This finding indicates a marginal height boundary between A and B ratings for marginal adaptation of 170 +/- 3 microm.