Depression in old age is underdiagnosed. In order to help detection, Yesavage et al. (1983) developed the Geriatric Depression Scale in a 30 items original version, with good reliability and validity. To abbreviate duration of examination, a shortened 15 items version was proposed (Sheikh et al., 1986) and a 4 items GDS for primary care attenders was also developed (D'Ath et al., 1994) with satisfactory metrological characteristics. The aim of this study is to develop a short French version and to contribute to its validation. French elderly depressives (181) and controls (66) were asked to complete the 30 items GDS (french version) and an other french self-rating scale, the QD2A (Pichot, 1986). In an attempt to devise the short scale, data were subjected to logistic regression analysis, multiple regression analysis, item-total Spearman's rank correlation coefficients and finally to mean choice ranks combination method. The generated 4 items version (mini-GDS) was found to be highly correlated with GDS 30 (r = 0.84, p < 0.0001) and with QD2A (r = 0.64, p < 0.0001) and had a high level of internal consistency (KR20 = 0.66). The sensitivity and specificity of the GDS 4 were 69% and 80% (cut-off 0/1) and against QD2A were 75% and 75% (cut-off 0/1). Out of the four items, two were common with the D'Ath's 4 items version. Method with its different statistical developments was discussed and the 4 selected items of the mini-GDS were compared with data from several authors who evaluated, discussed or examined factor structures of the 30 items GDS. Conclusion is that mini-GDS is reliable to easily detect a depressive symptomatology in the elderly, in different circumstances.