Objective - To assess the efficacy of spinal manipulation for patients with back or neck pain. Design - Computer aided search for published papers and blinded assessment of the methods of the studies. Subjects - 35 randomised clinical trials comparing spinal manipulation with other treatments. Main outcome measures - Score for quality of methods (based on four main categories: study population, interventions, measurement of effect, and data presentation and analysis) and main conclusion of author(s) with regard to spinal manipulation. Results - No trial scored 60 or more points (maximum score 100) suggesting that most were of poor quality. Eighteen studies (51%) showed favourable results for manipulation. In addition, five studies (14%) reported positive results in one or more subgroups. Of the four studies with 50-60 points, one reported that manipulation was better, two reported that manipulation was better in only a subgroup, and one reported that manipulation was no better or worse than reference treatment. Eight trials attempted to compare manipulation with some placebo, with inconsistent results. Conclusions - Although some results are promising, the efficacy of manipulation has not been convincingly shown. Further trials are needed, but much more attention should be paid to the methods of study.