We aimed to develop a better understanding and method of rating the success or failure of low back surgery by studying 185 patients prospectively. Identical pre-operative and postoperative assessment by an independent observer included pain, disability, physical impairment, psychological distress and illness behaviour. Outcome was assessed by the patient, by the observer and by return to work. There was 96% follow-up at two years. Correlation co-efficients varied considerably between the various measures of outcome, both patient and observer appearing to base their assessment mainly on postoperative status rather than on any change produced by surgery. The observer was influenced most by postoperative pain, disability and physical impairment. Patients were influenced most by residual physical impairment, type of surgery and proportional change in disability. Return to work was moderately influenced by postoperative disability and to a larger extent by social and work-related factors. We developed a simple formula to judge overall success or failure which accurately reproduced the combined assessment of patient and observer. If surgical audit is to be meaningful it must be based on an improved understanding of how the outcome of surgery should be assessed.