Many controversial areas exist in the treatment of spinal deformities in patients with cerebral palsy or mental retardation, including the benefits of surgery, the use of traction for preoperative curve correction, the need for a combined anterior and posterior approach, the need to fuse to the sacrum, and the moral question of operating on these severely handicapped patients. To help to clarify these questions, the surgical treatment of spinal deformities in patients with cerebral palsy or mental retardation was analyzed in 109 patients who were treated from May 1948 through December 1979 at the Twin Cities Scoliosis Center, Minneapolis and St. Paul, Minnesota. Seventy-seven (71 per cent) of the patients had cerebral palsy and thirty-two (29 per cent) had only mental retardation. One patient had lordosis only and one had kyphosis only. Of the 107 patients with scoliosis, forty-four had Group-I (double balanced) curves and sixty-three had Group-II (large unbalanced lumbar or thoracolumbar) curves. The incidence of the two curve types was found to differ in those patients with only mental retardation, in ambulatory patients, in patients who lived at home, and in patients with pelvic obliquity. The treatment programs in use during the thirty-one years that are covered by this study were: cast correction and posterior fusion followed by a long post-operative supine period; posterior fusion and Harrington instrumentation; and a two-stage combined anterior and posterior fusion and instrumentation with a very short postoperative supine period. The indications for surgery were curve progression (63 per cent), loss of function (35 per cent) and the magnitude of the curve (77 per cent). Traction was found to be of no use for correcting the curve, but was very useful in controlling the uncooperative patient. The length of follow-up averaged 4.5 years (range, two to twenty-nine years). All but ten of the patients achieved a solid spine fusion. Eight of the ten had painless pseudarthroses without loss of correction and two had pseudarthroses with loss of correction. The Group-II curves were better treated by the two-stage combined approach, which gave better correction of the scoliosis and a lower rate of pseudarthrosis compared with posterior fusion and instrumentation alone. The improvement in the results using the combined approach caused us to use this approach also in selected Group-I curves in the presence of a significant lumbar component. Fusion to the sacrum was necessary only when pelvic obliquity was present or sitting balance was absent. One patient was functionally worse post-operatively, eighty-two showed no change, and twenty-four showed improvement. The complication rate was high (81 per cent). The most frequent complications were pressure scores, wound problems, instrumentation problems, and an increase in the length of the curve. Pseudarthroses occurred in 17 per cent and infection, in 5 per cent of the patients. Three patients died and one became paraplegic. In our opinion, surgery can be of benefit in this group of severely handicapped patients.