Although laboratory studies indicate that acupuncture can have physiological effects, clinical studies of acupuncture are frequently inconclusive. Randomized controlled trials are the best way to test for the specific effectiveness of a therapy. Problems in clinical acupuncture research include the diversity of forms of therapy, individualised treatments, binding, choosing a credible control procedure, selecting suitable endpoints and traditional diagnosis. The enthusiasm of proponents of acupuncture may bias the research they are performing. None of these difficulties is insurmountable. Examples of rigorous trials are listed and some suggestions are offered on ways of improving research.