OBJECTIVE To compare the biomechanics of costotransverse process screw fixation with those of pedicle screw fixation in a cadaveric model of the upper thoracic spine. METHODS Ten human thoracic spines were instrumented across the T3-T4 segment with costotransverse and pedicle screws. Nonconstraining pure moments (maximum, 6.0 Nm) were applied to induce flexion, extension, lateral bending, and axial rotation. The range of motion, lax zone, and stiff zone were determined in each specimen in the normal state, after 3-column destabilization, and after instrumentation. After flexibility testing was completed, axial screw pull-out strength was assessed. RESULTS In all directions of loading, both fixation techniques significantly decreased lax zone and range of motion at T3-T4 compared with the destabilized state (P < .001). During all loading modes except lateral bending, pedicle screw fixation allowed significantly less range of motion than costotransverse screw fixation. Pedicle screws provided 62% greater resistance to axial pull-out than costotransverse screws. CONCLUSIONS The costotransverse screw technique seems to provide only moderately stiff fixation of the destabilized thoracic spine. Pedicle screw fixation seems to have more favorable biomechanical properties. These data suggest that the costotransverse process construct is better used as a salvage procedure rather than as a primary fixation strategy.