The purpose of the present study was to decide whether composite resin or conventional glass ionomer cement should be preferred as a base material in endodontically treated premolars. Twelve extracted human maxillary premolars were mounted in a universal testing machine at a 35 degrees angle. Cuspal stiffness was determined by applying a load of 75 N to the buccal cusp and recording the displacement of the cusp using inductive displacement transducers. In the same teeth, different cavity preparations and restorations were performed sequentially. Standard MOD cavities were enlarged to allow endodontic access. In addition, the cusps were undermined. Half of the teeth were restored to the level of the previous shallow cavities using conventional glass ionomer cement (Ketac Fil), in the rest of the teeth dentine bonding agent (Syntac) and composite resin (Tetric) were used instead. Finally, composite resin fillings (Tetric) were placed. All restorations were removed and the experiments were repeated twice. For each replication, the assignment of the base materials to the experimental groups was reversed, and ceramic inlays (Empress) were used as final restorations for the last replication. Improvement of cuspal stiffness achieved by conventional glass ionomer bases was very small, whereas composite resin bases increased cuspal stability by more than a factor of two. After placement of the final restorations, however, there was no longer a difference between teeth with different base materials. Nevertheless, composite resin bases might be preferred for two reasons. Firstly, deterioration of adhesive restorations will probably start at the cavosurface margins. The incidence of margin gaps, however, will not only compromise marginal seal but also the stabilizing effect of the restoration. In this situation, the resin base may still stabilize the tooth. Moreover, resin bases may reduce the risk of cusp fracture during the time between cavity preparation and the insertion of adhesive inlays.