The objective of this study was to visualize and to measure gap formation between dentin and restorative materials. The cusps of 15 extracted human sound molars were cut to expose fresh dentin. The teeth were randomized and divided into 3 groups (n = 5). Cubic cavities (6 x 4 x 4 mm3) were prepared with a diamond bur. The cavosurfaces were etched for 15 s. In order to induce adhesive failures eugenol was applied on one cavity wall, whereas the other cavity walls were left untouched. The cavities were filled with composite (Spectrum), compomer (Dyract AP) or ormocer (Admira) using the recommended bonding agent. All specimens were subjected to a thermomechanical cycling process of 1000 stress cycles (0 / 100N) as well as to 1000 temperature cycles (5° C / 55° C). The teeth were stored in saline at 20° C. Interfaces between cavity walls and restorative materials were inspected after 1 d, 14 d, 63 d and 144 d by confocal laser scanning microscopy (CLSM) in reflection mode. The gap widths (10 per tooth) were documented. In all specimens, measurable gaps could be found only at the eugenol treated surfaces. After 1 d the gap widths were 37.8 ± 4.6 µm for the composite, 30.2 ± 9.9 µm for the compomer and 34.4 ± 9.1 µm for the ormocer. Statistical evaluation (ANOVA) showed a significant (p Schlagwörter: microleakage, composite restoration, base material