The Beck Depression Inventory (BDI) and the General Health Questionnaire (GHQ) are commonly used in population studies as measures of depression. We examined in a population sample the validity of four scales for depressive symptoms, the GHQ-12, the 21- and 13-item versions of the BDI, and a new 6-item version of the BDI developed for this study. A total of 5561 participants in the "Health 2000" survey (30-79 years) completed the four scales and were assessed with the Composite International Diagnostic Interview (CIDI), which was used as the validation criterion. We selected items for the BDI-6 through an exploratory factor analysis for the BDI-21. The accuracy of the scales, including the BDI-6, was satisfactory (c-statistics 0.88-0.92 for depression within the past 2 weeks and 0.80-0.83 within the past 12 months) and slightly better for men (0.92-0.96 and 0.85-0.87) than for women (0.86-0.88 and 0.78-0.79). Higher scores in all the scales were associated with more severe depression and more recent depressive episodes. This study suggests that various versions of the BDI and the GHQ-12 are useful in detecting depressive disorders in the general population. Even the 6-item version of the BDI showed acceptable criterion validity, although replication in an independent dataset is needed to confirm its validity.