Scoliosis, lateral curvature of the spine, has been studied since Hippocrates' time, but remains a disputed subject in orthopaedic surgery, because of its several varieties, unknown cause and unpredictable course. A review of 30 years' experience in a paediatric orthopaedic unit was undertaken to clarify the problem. Patient records, collected prospectively, were examined to demonstrate the incidence, prevalence, extent, course and outcome of the commonest variety, adolescent idiopathic scoliosis. Reference was made to the results of previously published studies. Records from a school screening programme showed that, while slight degrees of spinal curvature were widely prevalent in the community, these were of no clinical significance and major cosmetic deformity was rare: only 8 in 10,000 adolescent girls had Cobb angles of 40 degrees or more and only half of these underwent surgical correction. Evidence to support non-operative treatment could not be demonstrated and it did not reduce the incidence of surgical intervention. Because the course of scoliosis did not seem to correspond with much published work, or with current hypotheses of aetiology, a rethinking of the whole subject is advocated. An alternative model of pathogenesis deriving from developmental biology was proposed. While advances in surgical methods have been significant, the core problems of aetiology and natural history remain. Until they are resolved, all conclusions on management must be provisional. This is where innovative thinking needs to be directed.