We retrospectively analyzed the postoperative neurological complications in 137 patients who underwent a posterior spine fusion for scoliosis and had concomitant somatosensory cortical evoked-potential spinal-cord monitoring. The patients were divided into three specific operative groups: group 1, forty-nine patients who had a Harrington rod with segmental wiring (segmental spinal instrumentation); group 2, twenty patients who had Luque segmental spinal instrumentation; and group 3, sixty-eight patients who had a Harrington rod without segmental spinal instrumentation. There were neurological complications in twelve (17 per cent) of the sixty-nine patients in groups 1 and 2. Three patients (4 per cent) had a major injury to the spinal cord and nine patients (13 per cent) had only transient sensory changes. No difference was apparent between group 1 and group 2 in the degree of operative correction of curves or in the incidence of neurological complications. The one neurological complication (1.5 per cent) that occurred in the sixty-eight patients in group 3 was a Brown-Sequard syndrome. The factors related to increased risk for spinal cord injury in groups 1 and 2 included: (1) the passage of sublaminar wires in the thoracic and thoracolumbar spine, (2) intraoperative correction exceeding the preoperative bending correction, and (3) the surgeon's lack of adequate experience with the technique. With spinal cord monitoring we were able to predict the impending major neurological deficits, but the transient (sensory) changes that may be associated with segmental wiring were less reliably predicted.