Purpose: To test the effect of different base materials on marginal quality of composite restorations after load cycling in vitro. Materials and Methods: Standardized Class I cavities were prepared in 110 extracted human molars. The teeth were randomized and divided into 10 groups (n = 11). Cavities of 8 groups were filled with composite using different base materials or cotton as negative control. Restorations of three more groups were performed with only a composite, a compomer or an ormocer, respectively. All specimens were subjected to a thermomechanical cycling process of 1000 stress cycles (0 / 100N) and 1000 temperature cycles (5° C / 55° C). After dye penetration with rhodamin, the teeth were sectioned and examined by conventional stereomicroscopy (SMi) and by confocal laser scanning microscopy (CLSM). Results: With both methods used, significant differences between the experimental groups were found (Kruskal-Wallis-Test: SMi p=0.02, CLSM p=0.001). Comparing the two microscopical methods it can be stated, that CLSM is more sensitive than SMi. Highest penetration values were detected at the negative control (mean rank: SMi 139, CLSM 161). Under SMi lowest values were found for the all compomer restoration (mean rank: 83) and for one restoration with a glasionomer used as base material (mean rank: 95). With CLSM lowest values were found for one glasionomer/composite (mean rank: 84) and for the all ormocer restorations (mean rank: 93). It is concluded that individual base materials might influence the marginal characteristics of Class I composite restorations. Keywords: microleakage, composite restoration