A number of treatments are widely prescribed for chronic back pain, but few have been rigorously evaluated. We examined the effectiveness of transcutaneous electrical nerve stimulation (TENS), a program of stretching exercises, or a combination of both for low back pain. Patients with chronic low back pain (median duration, 4.1 years) were randomly assigned to receive daily treatment with TENS (n = 36), sham TENS (n = 36), TENS plus a program of exercises (n = 37), or sham TENS plus exercises (n = 36). After one month no clinically or statistically significant treatment effect of TENS was found on any of 11 indicators of outcome measuring pain, function, and back flexion; there was no interactive effect of TENS with exercise. Overall improvement in pain indicators was 47 percent with TENS and 42 percent with sham TENS (P not significant). The 95 percent confidence intervals for group differences excluded a major clinical benefit of TENS for most outcomes. By contrast, after one month patients in the exercise groups had significant improvement in self-rated pain scores, reduction in the frequency of pain, and greater levels of activity as compared with patients in the groups that did not exercise. The mean reported improvement in pain scores was 52 percent in the exercise groups and 37 percent in the nonexercise groups (P = 0.02). Two months after the active intervention, however, most patients had discontinued the exercises, and the initial improvements were gone. We conclude that for patients with chronic low back pain, treatment with TENS is no more effective than treatment with a placebo, and TENS adds no apparent benefit to that of exercise alone.