Conservative therapies for low back pain (LBP) entail expense, work loss, and risk of side effects. Because many competing modalities have been advocated, 59 therapeutic trials were examined for adherence to 11 methodological criteria. Common problems included failure to randomize subjects, use "blind" observers, measure compliance, and adequately describe co-interventions. Applicability of many studies was unclear because of inadequate descriptions of patients, interventions, and relevant outcomes. Flexion exercises, administration of each of three drugs, one traction method, and certain manipulations were each supported by single studies of reasonable validity, but the importance of the results and their applicability to particular types of LBP were unclear. Valid trials supporting use of corsets, bed rest, transcutaneous nerve stimulation, and conventional traction were not found. Better methodological rigor is possible with newer techniques for ensuring blindness to therapy, measuring compliance, and assessing outcomes.