The objectives of this study were to determine the ability of the 30-, 15- and 8-item versions of the GDS for screening and assessing change in severity of depression in nursing home patients. The GDS and the MADRS were administered to 350 elderly NH-patients by trained interviewers. The presence of major (MaD) or minor depression (MinD) was evaluated with the Schedules for Clinical Assessment in Neuropsychiatry. Receiver Operator Characteristic (ROC) curves of the GDS-versions were performed to measure the ability to screen on depression. The ability to measure change in severity of depression was measured by differences in mean GDS-scores and mean MADRS-scores between patients with MaD, MinD and no depression, and expressed in terms of effect sizes. It was found that in ROC-curves all three GDS-versions performed well. The MADRS showed larger effect sizes for the differences between MaD, MinD and no depression than the GDS-versions. The effect sizes of the three GDS versions were comparable. We conclude that all three versions of the GDS can be used for screening on depression among NH-patients. The MADRS is superior to the GDS for assessment of (changes in) severity of depression, but the GDS also appears to be an acceptable instrument for this purpose and is less time-consuming.