A three-surface MOD preparation measuring one third of the isthmus cavity distance at the isthmus was prepared for five extracted, intact human maxillary premolar teeth. The teeth were restored with a posterior resin and a dentin bonding agent. Each tooth was tested under a cyclically varying load in a fatigue machine made at the University of Washington. The cyclic load rate was 75 cycles per minute. A strain gauge was applied to the buccal surface of each tooth. Under the load applied at an angle of 60 degrees to the long axis of the tooth, reinforcement of the tooth was established by the reduced deflection of the buccal surface of the restored tooth as compared with the unrestored tooth. The loss of this reinforcement due to the application of the cyclically varying load established the end point of the fatigue testing. Four different load values of 13, 16, 19, and 22 lb loads were applied separately to each tooth. The results of this testing were classic wherein the smallest load required the greatest number of load cycles before loss of adhesion was registered by the strain gauge. Failure always occurred within the adhesive joint.