The authors reviewed 246 idiopathic scoliosis patients with the upper thoracic curve of more than 20 degrees. Group I (138 patients) had positive T1 tilt and a spinal fusion which was extended over both the upper and lower thoracic curve with the diagnosis of double thoracic curve. Group II (43 patients) had positive T1 tilt, but the fusion was limited to the lower thoracic curve. Group III (65 patients) had negative or neutral T1 tilt and the fusion was limited to the lower thoracic curve. The average age at operation was 15.9 years (range, 11.2-35 years) and the average length of follow-up was 4.8 years (range, 2-29.5 years). Positive T1 tilt did not correlate well with left shoulder elevation contrary to previous reports. The upper thoracic curve was more rigid than the lower curve in all groups and the lumbar curve was significantly more flexible than the upper and lower thoracic curves in all groups (P < 0.05). No significant difference in the flexibility of the upper thoracic curve was found between the groups regardless of the direction of T1 tilt. When only the lower curve was fused (groups II and III), progression of the upper thoracic curve was less than 5 degrees, and spontaneous correction of the unfused upper curve occurred in the majority of the cases following the supine bending study. Correction and fusion on the lower curve (groups II and III) aggravated shoulder imbalance of all patients with left shoulder elevation. Based on the findings of this study, the authors proposed that the diagnosis of idiopathic double thoracic patterns should be limited to those patterns which require fusion of both the upper and lower curves. This pattern of idiopathic scoliosis includes double thoracic curves with left shoulder elevation and/or a rigid upper thoracic curve.