We put forward the hypothesis that general practitioners (GPs) with a family medicine orientation are more sensitive to the presence of mental health problems than GPs with a clinical orientation. To test it, GPs were divided into three subgroups on the basis of an attitude questionnaire. The General Health Questionnaire (GHQ) and the Present State Examination (PSE) were used as criteria. No differences in sensitivity to psychiatric illness were observed using either scale. Results of factor analysis with the subscales of the GP attitude questionnaire and the indices 'bias' and 'accuracy' were similar to those reported by GOLDBERG and associates. 'New' patients were defined as patients in whom the GP had not identified a mental health problem (MHP) in the past year. 'Old' patients were defined as 'not new'. GPs tended to under-identify MHPs in 'new' and over-identify them in 'old' patients. Recognition of psychiatric illness was better in 'old' than in 'new' patients. In 'new' patients, recognition depended on psychiatric diagnosis. Among 'old' patients, older people and people (especially women) with low education predominated. In their assessments GPs used information not contained in the GHQ.