STATEMENT OF PROBLEM: Cyclic thermal fatigue has demonstrated a significant influence of the thicknesses of luting composite and ceramic in crack propensity of porcelain laminates. PURPOSE: This study was conducted to define potentially involved parameters for crack development in porcelain laminates bonded to teeth. Finite element modeling was used to evaluate the respective effects of luting composite shrinkage and significant thermal changes. MATERIAL AND METHODS: A buccolingual cross-section of a maxillary incisor was digitized and used as a template to generate a single 2-dimensional mesh, including all the different restorative designs. Luting composite shrinkage was simulated at a baseline temperature of 20 degrees C. The effect of thermal loads from 20 degrees C to 5 degrees C and from 20 degrees C to 50 degrees C was assessed with and without preexisting composite shrinkage. RESULTS: Shrinkage of the luting composite alone generated important compressive forces on the ceramic, either at the restoration surface or interface. Compression intensity was related to geometry and ratio of thicknesses between the ceramic and luting composite (CER/CPR). Lower ratios produced higher compression forces in the ceramic. When thermal loads were combined to the composite shrinking forces, the stress pattern was significantly changed only for the experimental conditions with the lowest CER/CPR ratio. Temperature increase reduced compressive stresses and exacerbated tensile stresses. Thermal loads were simulated alone (situation of an "ideal nonshrinking" luting composite) and generated mainly tensile stresses in the ceramic, which intensity was again modulated by the CER/CPR ratio and the local geometry of composite and ceramic. Because of ceramic brittleness, these tensile forces were more detrimental than the high compression created by composite shrinkage alone. The stress pattern was not influenced by the incisal length of the veneer but rather by the facial thickness of ceramic. The worst record made with a shrinking luting agent (500 microm of luting composite, lowest CER/CPR ratio, 5 degrees C) was much less harmful than the worst record made with a hypothetical "nonshrinking" luting material. CONCLUSIONS: The ratio of the thickness of cement and luting composite appears to have a relevant influence on the stress distribution in porcelain laminates. Restorations that are too thin, combined with poor internal fit, resulted in higher stresses at both the surface and interface of the restoration. Because of its precompressed state given by composite shrinkage, ceramics performed better with regard to temperature-induced tensile forces.