We analyzed the results of posterior arthrodesis of the spine for congenital scoliosis, with or without Harrington instrumentation, in 290 of 323 patients who were operated on between the ages of five and nineteen years and were followed for two years or more. The length of follow-up averaged six years and ranged from two to twenty-eight years. The average curve before surgery was 55 degrees (range, 13 to 155 degrees), the average curve at correction was 38 degrees (range, 5 to 102 degrees), and the average curve at final follow-up was 44 degrees (range, 5 to 103 degrees). Bending of the fusion mass of more than 10 degrees was seen in forty patients; pseudarthrosis, in twenty; and adding-on of vertebrae with an increase in the curve of more than 10 degrees, in seven patients. There were four deaths, only one of them in the last twenty-five years. One was due to intraoperative cardiac arrest; one, to intraoperative overtransfusion; one, to postoperative overtransfusion; and one, to gastrointestinal bleeding eight months postoperatively while the patient was in a Risser jacket. Two patients became paraplegic due to excessive distraction with the Harrington rod, and two others had a partial cranial-nerve lesion due to halo traction. Based on these results, we concluded that posterior arthrodesis of the spine is satisfactory for most patients with congenital scoliosis. The most common problem was bending of the fusion mass in growing children, which occurred in 14 per cent of the patients. Use of Harrington instruments allowed slightly better correction (36 per cent compared with 28 per cent) but was associated with the only cases of paraplegia and infection in the series.