Fifty-eight experimental resin restorations were performed in intact, human premolars, using different leakage-reducing restorative procedures. These were conventional acid-etching and acid-etching followed by cavity treatment with an intermediary layer of low-viscous resin or the dentin adhesive NPG-GMA/ethanol. The teeth were extracted after 4 months and examined for pulpal inflammation/necrosis (I), reduction of odontoblasts (OR), and formation of tertiary dentin (TD). By the general linear model procedure, 91%, 34%, and 56% of the variations in I, OR, and TD, respectively, could be explained by variations in the experimental conditions. The significant independent variables were jaw, stage of root formation, width of pulp, width of cavity, marginal leakage, bacteria in the cavity, bacteria in the exposed dentinal tubules, and the restorative procedure. With regard to the restorative procedure the analyses showed that application of low-viscous resin increased the pulpal reactions OR and I, whereas cavity treatment with NPG-GMA/ethanol had no adverse biologic effect.