BACKGROUND: Diagnosis of depressive disorder using interviewer-administered instruments is expensive and frequently impractical in large epidemiological surveys. The aim of this study was to assess the validity of three self-completion measures of depressive disorder and other psychiatric disorders in older people against an interviewer-administered instrument. Method A random sample stratified by sex, age and social position was selected from the Whitehall II study participants. This sample was supplemented by inclusion of depressed Whitehall II participants. Depressive disorder and other mental disorders were assessed by the interviewer-administered structured revised Clinical Interview Schedule (CIS-R) in 277 participants aged 58-80 years. Participants also completed a computerized self-completion version of the CIS-R in addition to the General Health Questionnaire (GHQ) and the Center for Epidemiologic Studies Depression Scale (CES-D). RESULTS: The mean total score was similar for the interviewer-administered (4.43) and self-completion (4.35) versions of the CIS-R [95% confidence interval (CI) for difference -0.31 to 0.16]. Differences were not related to sex, age, social position or presence of chronic physical illness. Sensitivity/specificity of self-completion CIS-R was 74%/98% for any mental disorder and 75%/98% for depressive episode. The corresponding figures were 86%/87% and 78%/83% for GHQ and 77%/89% and 89%/86% for CES-D. CONCLUSIONS: The self-completion computerized version of the CIS-R is feasible and has good validity as a measure of any mental disorder and depression in people aged 60 years. GHQ and CES-D also have good criterion validity as measures of any mental disorder and depressive disorder respectively.