In this longitudinal study (12 and 24 months follow-up) of 216 patients with early rheumatoid arthritis (RA) (< or = 4 yr duration), we assessed the strength and stability of the relationship between psychological distress and traditional clinical variables examined the ability of these clinical variables to predict changes in mental distress, and explored the directionality between mental distress and the other clinical variables. Study variables were symptoms of anxiety and depression measured by the Arthritis Impact Measurement Scales, tender joint counts, erythrocyte sedimentation rate (ESR), and self-reported pain and disability. Psychological distress showed high levels of stability over time. Pain and disability were the two variables most strongly and consistently related to mental distress. High levels of disability predicted an increase in depression during the next year; otherwise, changes in psychological distress were not predicted by disease-related variables. Attempts at causal modelling of the temporal relationship between mental distress, pain and disability failed to yield consistent results.