For improving the marginal integrity of composite restorations, a total bonding method has been advocated besides the conventional selective bonding method. Total bonding avoids the placement of a base. The entire internal cavity surface is available for the adhesive bond. Selective bonding involves the placement of a base covering the pulpal floor as well as the pulpoaxial wall. In this study, five dentin/enamel bonding system/composite combinations were used to restore 60 class 2 cavities with their cervical margins below the cementoenamel junction (CEJ); six teeth per material were restored according to the total bonding and the selective bonding methods. Before and after simultaneous thermo-cycling and mechanical loading (TCML) marginal adaptation was evaluated on replicas in the SEM. Microleakage was determined by dye penetration on the original samples after TCML. The data were statistically evaluated with the Mann-Whitney U-test and the Wilcoxon test. The error rates method was applied. In SEM analysis the error rates method indicated a significant difference between the two restoration methods in general. In the pairwise comparisons, no significant differences between the selective bonding and total bonding methods were found for Syntac/Tetric, Gluma 2000/Pekafill, and Gluma/Pekafill. With Scotchbond Multi-Purpose (SBMP) and All-Bond 2 (AB2), total bonding revealed significantly (P < or = 0.01) less gap formation before and after TCML than selective bonding. Accordingly, total bonding showed significantly less (P < or = 0.01) dye penetration with Scotch-bond Multi-Purpose and All-Bond 2 compared to selective bonding. In conclusion, the reduction of microleakage by application of the total bonding method depended upon the bonding system used. Total bonding could be an alternative procedure for the adhesive restoration of class 2 cavities when their gingival margins are apical to the cementoenamel junction, provided the proper system is used and pulp damage is prevented.