An observation method for the assessment of pain behaviors in patients with rheumatoid arthritis (RA) has been developed. We investigated the extent to which the frequencies of pain behaviors differentiated patients with RA and patients with chronic low back pain from depressed and nondepressed, pain free, control subjects. The reliability of the pain behavior frequencies of patients with RA across 2 observation sessions also was determined. Total pain behavior scores clearly differentiated patients with RA and low back pain from depressed and nondepressed, pain free, control subjects. Pain behavior observed in patients with RA showed a high degree of stability over time. The results of our study suggest that the behavioral observation method will prove useful in the assessment of RA pain in clinical and research settings.