Anterior, posterior, and combined thoracolumbar spinal injuries were simulated in human cadaver specimens and then stressed in flexion, before and after stabilization, with Weiss Springs, Roy-Camille plates, vertebral body plates, and Harrington compression and distraction rods. The reduction, stability, and strength were determined for each implant-injury combination. For posterior ligamentous injuries, the Harrington compression system on the laminae gave reduction in extension, stability to that of the intact spine, and failure at 87.6 Nm bending load with 14.1 J energy absorption. For anterior vertebral body fracture, Harrington distraction rods from three vertebrae above to three below the injury gave a reduction in extension with stability similar to that of the intact spine. Failure occurred at 81.6 Nm load, one-third greater than with rods two levels above to two below, and 14.0 J energy absorption, twice that for the short rod. The more unstable combined anterior and posterior injury was satisfactorily reduced only by the long distraction system, which failed at 44.1 Nm load, twice that for the short rod, and 5.7 J energy absorption. By accurately determining what structures have been injured, and appropriate fixation device can be selected and the strength of the stabilized spine estimated. A thorough understanding of the biomechanics of the spine is essential for successful clinical utilization of these experimental data.