We compared the effects of four treatment variables on the pain reduction produced by transcutaneous electrical nerve stimulation and attempted to establish indications for TENS based on patient history and pain evaluation items. Treatment variables were the therapist and the three TENS stimulus characteristics--pulse width, frequency, and amplitude. We randomly assigned 192 consecutive adult patients suffering from painful conditions to one of four physical therapists and one of 12 stimulus characteristic combinations. We used a standard evaluation form that included a visual analog scale (VAS) to evaluate pain. Patients were given a 30-minute trial using TENS, followed by reevaluation. The VAS line length change after treatment was the criterion score for comparison of stimulus characteristic effectiveness. Although pain was reduced greatly with TENS (p = .01), a four-way analysis of variance (pulse width, frequency, amplitude, and therapist) attributed little of the treatment effect to the treatment variables or their interactions (r2 = .101). The amplitude effect, however, was borderline (p = .056), and subthreshold stimulation proved more effective than stimulation to tolerance (p = .05). Extensive multiple linear regression analyses failed to provide indications for TENS based on patient information and pain evaluation items. Therefore, pain remains the only indication for TENS, and we recommend subthreshold rather than higher amplitude stimulation on the initial TENS trial.