We have investigated the affect of removal of Harrington rods in 36 patients who had undergone successful spinal fusion for idiopathic scoliosis. In the curves in which further deterioration had taken place even though fusion had been achieved, there was a small improvement in both major and minor curves in the first six months after removal of the rod, followed by complete stabilisation of the curve. A similar tendency to stabilisation was noticed in the curves which had not shown deterioration. We conclude that regression of a curve which has fused following Harrington rod instrumentation occurs when the rod has not been placed according to optimum biomechanical principles. Removal of the rod then relieves the deforming stresses and allows the spine to stabilise.