In a detailed physical and psychiatric assessment of 80 patients with definite or classical rheumatoid arthritis (RA) different instruments were used to measure psychiatric disorder. The prevalence of psychiatric disorder was 21% when assessed by the PSE/CATEGO programme and 24% according to RDC criteria, but these figures were nearly doubled if a lower threshold was used to define psychiatric disorder. This study demonstrates how symptoms directly attributable to arthritis may inflate the estimated prevalence of psychiatric disorder in RA and erroneously indicate a direct relationship between severity if RA and psychiatric disorder. In fact, the best prediction of psychiatric disorder resulted from using a combination of measures of social stress and severity of RA.