Objective The CES-D is an instrument that is commonly used to screen for depression in community-based studies of the elderly, but the characteristics of the CES-D in a residential home population have not yet been studied. The aim of this study was to investigate the criterion validity and the predictive power of the CES-D for both depressive and anxiety disorders in a vulnerable, very old population living in residential homes. Methods Two hundred seventy seven residents were screened with the CES-D, and subsequently interviewed with a diagnostic instrument, the Mini International Neuropsychiatric Instrument (MINI). The sensitivity, specificity, and positive and negative predictive value of the CES-D were calculated by cross-tabulation at different cut-off scores. Receiver Operating Characteristics (ROC) curves were used to assess the optimal cut-off point for each disorder and to asses the predictive power of the instrument. Results In a residential home population the CES-D had satisfactory criterion validity for depressive disorders and for any combination of depressive and/or anxiety disorders. With a desired sensitivity of at least 80%, the optimal cut-off scores varied between 18 and 22. The predictive power of the CES-D in this population was best for major depression and dysthymia (Area Under the Curve, AUC 0.87), closely followed by the score for any combination of depressive and/or anxiety disorder (AUC 0.86). Conclusion The use of one single instrument to screen for both depression and anxiety disorders at the same time has obvious advantages in this very old population. The CES-D seems to be a suitable instrument for this purpose. Copyright © 2010 John Wiley & Sons, Ltd.