Physicians were surveyed regarding their beliefs about treatment efficacy for patients with low back pain. To document physician beliefs about the efficacy of specific treatments and the extent to which these beliefs correspond to current knowledge. Little is known about physician beliefs regarding the efficacy of specific back pain treatments. A national random sample of 2897 physicians were mailed questionnaires that asked about 1) the treatments they would order for hypothetical patients with low back pain and 2) the treatments they believed were effective for back pain. Responses were compared with guidelines suggested by the Quebec Task Force on Spinal Disorders. Almost 1200 physicians responded. More than 80% of these physicians believed physical therapy is effective, but this consensus was lacking for other treatments. Fewer than half of the physicians believed that spinal manipulation is effective for acute or chronic back pain or that epidural steroid injections, traction, and corsets are effective for acute back pain. Bed rest and narcotic analgesics were recommended by substantial minorities of physicians for patients with chronic pain. The Quebec Task Force found little scientific support for the effectiveness of most of the treatments found to be in common use. The lack of consensus among physicians could be attributable to the absence of clear evidence-based clinical guidelines, ignorance or rejection of existing scientific evidence, excessive commitment to a particular mode of therapy, or a tendency to discount the efficacy of competing treatments.