The effect of variations in the design of class-II cavity preparations on the marginal degradation of amalgam restorations was included as a study aim in a clinical trial. Four hundred and sixty-eight restorations were placed in 210 patients by 7 Scandinavian dentists using 5 different alloys. The marginal degradation was scored on impressions of the restored teeth by means of a six-point ordinal rating scale. The scores were then compared with defined characteristics of the occlusal parts of the initial cavity preparations. Characteristics of the cavity that could be related to the marginal degradation were diverging occlusal cavity walls, occlusal cavity depth, fissures perpendicular to the cavosurface angle, and rough or variable occlusal cavosurface angles. Cavity preparation features not influencing the rate of degradation were the occlusal width, the location of the cavosurface angle on the cusp slope, occlusal cavosurface angles with sectors smaller than 90 degrees, and less than 1 mm enamel remaining between the cavity preparation and another restoration. The association between the different cavity design features and the marginal degradation varied with the different alloys. Superior marginal performance is probably the result of optimal condensation or surface treatment, rather than features of the cavity preparation.