Objective. To test the reliability and validity of a self-administered 36 joint count developed after the Rapid Assessment of Disease Activity in Rheumatology (RADAR) questionnaire for assessing pain/tenderness. Methods. Two self-administered formats (mannequin and text) were evaluated in 60 patients with rheumatoid arthritis (RA). Reliability between both formats was tested by Spearman rank correlation. Criterion validity/accuracy was tested by Spearman correlation coefficient between each self-report format and a joint count performed by a physician. Construct validity was ascertained by correlation of each format with other variables of disease activity. Results. Reliability between the 2 formats was high (R = 0.94). Correlations between each format and the physician's joint count were also high (R = 0.77 for mannequin, 0.75 for text). Patients consistently rated their joint pain/tenderness higher than the physician (means 29, 27, and 12 for text, mannequin, and physician, respectively; p < 0.01). Construct validity of the text, mannequin, and physician formats compared with the modified Health Assessment Questionnaire showed R = 0.61, 0.65, 0.63; with Steinbrocker functional class R = 0.41, 0.46, 0.56; with pain R = 0.59, 0.61, 0.62; with global evaluation R = 0.66, 0.71, 0.84; and with morning stiffness R = 0.64, 0.59, 0.60, respectively. Conclusion. Although both self-administered formats exhibited adequate reliability and construct validity, a systematic difference between patient and physician/trained assistant performed joint counts was observed, with patients consistently rating their pain/tenderness higher. We thus do not believe they can replace standard physician/trained assistant evaluation in obtaining clinical research data in rheumatology.