Therapeutic trials often attempt to "blind" patient and investigator to the true nature of treatments received, reducing the influences of conscious or subconscious prejudices. In drug trials, this is accomplished with placebo tablets, but blinding in trials of physical treatments is more problematic. This issue arose in a clinical trial of transcutaneous electrical nerve stimulation (TENS) for patients with chronic low back pain. Several study design features were incorporated to promote blinding: use of sham TENS units visually identical with real units, exclusion of potential subjects with previous TENS experience, avoidance of a crossover design and use of identical visit frequency, instructions and modifications in electrode placement. Subjects were asked not to discuss treatments with the clinicians who performed outcome assessments. Both patients and clinicians were asked to guess actual treatment assignments at the trial's end. Every patient in the true TENS group believed the unit was functioning properly, but the degree of certainty varied. In the sham TENS group, 84% also believed they had functioning units, but their certainty was significantly less than in the active treatment group. Differences in patient perceptions did not affect compliance, as the two groups had similar dropout rates, appointment compliance, days of TENS use and daily duration of TENS use. Clinicians guessed treatments correctly 61% of the time (as opposed to 50% expected by chance), again suggesting partial success in blinding. These efforts at blinding may partly explain the negative trial results for TENS efficacy. We conclude that complete blinding is difficult to achieve because of sensory difference in treatment and unintended communication between patient and examiner.(ABSTRACT TRUNCATED AT 250 WORDS)