A total of 40 extracted human maxillary premolar teeth were used in this study. Teeth were randomly distributed between two main groups. In group A conservative class II Box preparations were cut in each tooth. In group B the proximal cavity was prepared through the occlusal surface and the proximal marginal ridge was left intact (tunnel preparation). Group A and group B were randomly subdivided into two subgroups (a & b). In groups Aa and Bb a posterior composite resin material was used to restore the prepared cavities (no base was used). While in groups Aa & Bb the posterior composite resin was used with a glass ionomer cement base. The teeth were tested on an Instron testing machine for their resistance to fracture, force being applied equally to buccal and lingual cusps. Mean fractures values were: Gp. Aa 285 LB.: Gr. Bb 245 lb. Bb 240 lb. One way analysis of variance showed no statistical difference between the four groups. The tunnel preparation did not increase the resistance to fracture of the restored tooth compared to a box preparation.