Finite element (FE) analysis of the mechanical behaviour of materials and structures facilitates the investigation of their internal stress distributions. However, the validity of the model is not always ascertained. In this study a three-dimensional (3D) FE model was developed, representing a laboratory set-up of direct post and core restored upper premolars. These restorations, using either composite or amalgam for core material, have been the subject of study in previous quasistatic and fatigue strength tests. The aim of this study was to validate the FE model for prefailure and failure modelling, by comparing the computational results with the laboratory observations and failure results. Two failure criteria were selected for investigation: Modified Von Mises and Drilcker-Prager equivalent stress. Four model variations were carried out, representing different conditions at the core-tooth interface. Prefailure modelling was found to be adequate. The calculated failure results could only partly be fitted to the quasistatic tests. The best fit was effected with a model using partial bonding of the core, for the composite core. Fatigue failure was reproduced somewhat better by a model using no bonding at all, again to a higher degree for the composite core. Calculations of post stress using a model simulating increased core mobility supported an observation made previously (M. C. D. N. J. M. Huysmans et al., in press; Int. Endodont. J. XX,XXX-XXX), implying that a composite core raises the demands made on the post. The conclusion is made that validation of FE calculations is essential. A 3D model as presented here shows a satisfactory fit to fatigue data but not to quasistatic results. Further development of suitable failure criteria is indispensible.