Previous reviews of psychological factors in arthritis have emphasised the methodological weaknesses of many studies, especially those attempting to measure personality after years of disabling disease. To make sense of the published reports three factors need to be considered separately: previous personality, social stresses, and current mental state. Each can now be measured reliably and independently of symptoms which might be directly attributable to the arthritis. There is a growing consensus that the normal range of personality is represented among patients with early arthritis, that the prevalence of depression is similar to that of patients with other medical conditions, and that social stress is more closely related to depression than activity and the disabling effect of arthritis. Longitudinal studies are now required to examine which social stresses can be attributed to the disabling effect of arthritis. Depression and social stress often manifest themselves to the rheumatologist as excessive complaints of pain and frequent clinic attendances so appropriate psychosocial treatments may reduce this behaviour.