Objective: To determine the diagnostic accuracy of the 30-item and shortened versions of the Geriatric Depression Scale (GDS) in diagnosing depression in older nursing home patients. Method: Three hundred and thirty-three older nursing home patients participated in a prospective cross-sectional study in the Netherlands. Sensitivity and specificity, positive and negative predictive values, and the area under the receiver operating curve (ROC) were assessed. Cronbach alphas were also calculated. Both major depression (MDD) and minor depression (MinD) according to the DSM-IV criteria, measured with the Schedules of Clinical Assessment in Neuropsychiatry (SCAN), were used as 'gold standard'. Results: The cut-off point≥ 11 on the GDS-30 gave a sensitivity of 96.3% for NMD and 85.1% for MinD, with a specificity of 69.1%. The sensitivity of most of the shortened versions was sufficient, varying between 88.9% and 100% for NMD, and between 63.8% and 97.9% for MinD. With regard to the shortened versions, best sensitivity (96.3% and 78.7%) and specificity (69.5%) were found for the GDS-10 developed by D'Ath et al. (1994). The specificity rates for most of the shortened versions were found to be less satisfactory, varying between 18.9% and 74.1%. Sufficient internal consistency was found for the GDS-30, the GDS-15, the GDS-12 and the GDS-10, with Cronbach's alphas varying between 0.88 and 0.72. Conclusions: The GDS-30 was found to be a valid and reliable case-finding tool for both major and minor depression in nursing home patients with no cognitive impairment and in patients with mild to moderate cognitively impairment (MMSE ≥ 15). The GDS-10 (D'Ath et al., 1994) appeared to be the best least time-consuming alternative for the nursing home setting.