BACKGROUND. Many scales have been developed to assess depression, but they are often too lengthy to be of practical use in general practice consultations. AIM. A study was undertaken to investigate the feasibility, reliability and diagnostic value of the geriatric depression scale and its shorter versions for screening in general practice. METHOD. A total of 586 consecutive consulting patients aged 65 years and over were studied in nine general practices in the west of the Netherlands (13 doctors). The 30-item version of the geriatric depression value was compared with the diagnostic interview schedule as a reference test. RESULTS. The reference test indicated a major depression in six patients while 27 patients had a dysthymic disorder (that is, a chronic mild depression). Five per cent of patients required help for 50% of the questions on the geriatric depression scale. The diagnostic value of the 30-item, 15-item, 10-item and four-item versions did not differ significantly, but the one-item version performed no better than chance. Two items discriminated best between patients who were and who were not depressed (P < 0.05), only one of which was included in a previously proposed four-item version of the scale. The reliability of the proposed four-item version was 0.64, the reliability of the other versions ranging from 0.70 to 0.87. CONCLUSION. The results for the different versions of the geriatric depression scale suggest the use of a 10-item or a four-item version. For practical purposes, the smallest subset would be the most desirable: the four-item version. These scales may be better suited for exclusion rather than inclusion purposes. The feasibility of screening for depression in elderly people in a general practice setting is discussed in the light of the results of the study.