The first-degree relatives of 50 obsessive-compulsive patients and those of matched controls completed the General Health Questionnaire (GHQ) and the Leyton Obsessional Inventory (LOI). Relatives who were identified as possible 'cases' by their high GHQ scores, or by their own or informant relatives' reports, were interviewed using the Schedule for Affective Disorders and Schizophrenia. Index relatives had a significantly higher lifetime prevalence of mental illness (36%) than had those of controls (17%), due mainly to an excess of depressive and neurotic disorders. However, only one relative from each group was diagnosed as having definite obsessive-compulsive neurosis. In addition, the LOI scores were similar for the index and control relatives.