Twenty-three fresh-frozen human specimens were subjected to ankle arthrodesis and fixation with two cancellous-bone screws. The specimens were then subjected to four newton-meters of manually applied tibial torque, plantar flexion-dorsiflexion moment, and medial-lateral bending moment; relative rotation between the tibia and the talus was recorded for each mode of testing. A Calandruccio triangular compression device was then applied to threaded pins penetrating the tibia and talus, the screws were removed, and the test sequence was repeated. Tibiotalar motions recorded with both systems of fixation were markedly affected by the quality of the bone. When less than 2 degrees of total tibiotalar rotation was recorded in response to four newton-meters of manually applied internal-external tibial torque, all specimens demonstrated less torsional rotation with the screw fixation than with the external fixator. The mean rotations produced by medial-lateral bending moment were equivalent for both systems of fixation. When torsional rotations with the use of screw fixation were greater than 2 degrees, all specimens demonstrated more torsional rotation with the screws than with the external fixator, and all but one specimen had more medial-lateral rotation with the screws than with the fixator. For applied plantar flexion-dorsiflexion moment, twenty-two of the twenty-three specimens demonstrated more rotation with the Calandruccio fixator than with the screws; this was due in part to motion permitted at the hinge points of the frame itself.