A group of 109 patients with unilateral low back pain for over three months were randomised to receive one of three types of injection treatment: cortisone and local anaesthetic injected into two facet joints (28), the same mixture around two facet joints (39), or physiological saline into two facet joints (42). The effect of the treatment was evaluated in relation to work attendance, pain, disability and movements of the lumbar spine. Patients were examined one hour and two and six weeks after treatment and also completed a questionnaire after three months. A significant improvement was observed in work attendance, pain and disability scores, but this was independent of the treatment given and movements of the lumbar spine were not improved. Of the 70 patients with initial pain relief after injection, 36% reported persisting benefit at the three month follow-up, independent of the mode of treatment given. We conclude that facet joint injection is a non-specific method of treatment and the good results depend on a tendency to spontaneous regression and to the psychosocial aspects of back pain.