Eleven cases of fracture or fracture/dislocation of the thoracolumbar or lumbar spine were studied to determine the effectiveness of three methods of management of the injuries. Ten of the patients had a neurological deficit as well as a spinal injury. All cases were studied by computed tomographic (CT) scanning of the injury site before and after treatment. CT scanning was found to be superior to linear tomography in determining the degree of canal compromise by bone at the injury site. Eight patients underwent operation, with the aims of achieving spinal stability in unstable injuries and of decompressing neural structures. These 8 patients had received Harrington distraction instrumentation (HDI) as an initial procedure. In those patients with fracture/dislocations from flexion-rotation injuries and canal compromise resulting from bony malalignment, reduction of the fracture/dislocation by HDI resulted in stabilization of the spine and increase of the open canal area at the injury site. However, in those patients with canal compromise resulting from bursting fractures with retropulsed bone fragments, HDI did not reduce the bone fragments. Removal of the anterior vertebral body and the bone fragments. however, did result in significant improvement of the canal dimensions. A poor correlation was found between the degree of canal compromise as measured by the CT scanner and the resulting neurological deficit. Reasons for this are presented. Six of the 10 neurologically compromised patients, including 2 patients with stable bursting injuries and retropulsed bone fragments treated conservatively, had improved one grade on Frankel's classification by 3 months after injury. The rate and degree of recovery were not related to the degree of canal decompression achieved at operation.