1. The study compares the psychometric performance of the CES-D and the GHQ-12 in a sample of elderly community residents. Misclassification rates of the questionnaires were analyzed and suggestions for improvement of scale performance are made. 2. 287 subjects out of the general population aged 60-99 years were personally interviewed with standardized diagnostic tools and completed both the GHQ-12 and the CES-D. Best-estimate diagnoses served as standards for receiver operating characteristics (ROC) analysis. 3. Both the GHQ-12 and the CES-D discriminated well between depressive and nondepressive subjects (AUROC = 0.794 and AUROC = 0.782, respectively). The amount of false positive results was high for both questionnaires (GHQ-12: 80.6%, CES-D: 90.1%). Increasing age led to more false positive results on the GHQ-12, whereas the CES-D yielded more false positive results in subjects living in an old age residence or together with family members when compared to those living together with their spouse. 4. The GHQ-12 and the CES-D were valid screening instruments for depression in a community sample of elderly subjects. However, both questionnaires yielded a considerable proportion of false positive results. Elevation of the cut-off score may reduce the misclassification rate of the GHQ-12 but not that of the CES-D.