A double-blind control study of lumbar traction for sciatica has been carried out. Although there is a tendency for traction to produce improvement in pain and straight-leg raise the extent does not achieve statistical significance. Changing ’control‘ patients to ’treatment‘ seemed to produce worthwhile relief of pain for all who were not already improving. It is suggested that a large trial using more discriminating criteria might delineate a group of patients susceptible to help by traction.