The Afar triple junction and the surrounding region provide a classic example of continental rifting preceeded by the impingement of a mantle plume. Volcanism began in the Oligocene with the emplacement of a massive flood basalt pile in Ethiopia and Yemen, and continues to the present day in the Red Sea and Gulf of Aden oceanic spreading centers and the East African continental rift. This volcanism is currently the subject of a collaborative French-Ethiopian project aimed at understanding the relationship between plume impingement and continental break-up. In the context of this project, we have undertaken an Os isotopic study of rift and flood basalts from Ethiopia and Djibouti. Because of the strong isotopic contrast between the mantle and the crust, Os isotopes are especially useful for constraining processes of crustal contamination. In particular, they allow assimilation of crustal material by basaltic magma to be distinguished from direct melting of metasomatized mantle lithosphere. Os isotopic ratios of rift and flood basalts are plotted as a function of 1/[Os] in Fig. 1. It is evident that very few of the rift basalts have Os isotopic values compatible with direct mantle derivation. There is a rough correlation between lSTOs/188Os and 1/[Os]. That is, rocks with lower Os concentrations (2 ppt, in the most extreme case), resulting from higher degrees of differentiation, are most sensitive to the effects of crustal contamination. Further information concerning crustal assimilation can be obtained by comparing Os and He isotopic data. Helium, though geochemically very different from Os, is also a highly sensitive indicator of crustal addition. He is also the only isotopic tracer capable of clearly indicating the presence of a deep, less degassed, mantle component. A direct comparison of Os and He isotopes in the rift basalts yields no correlation. This is because the 187Os/188Os ratio is controlled by the Os concentration of the rock as well as by the total amount of contamination. On the other hand, a negative correlation (Fig. 2) is observed between the proportions of crustal assimilation required to explain the Os isotopic compositions, given the measured Os concentrations, and the 3He/4He ratios of the rift samples. (Decreasing 3He/4He implies increasing crustal contamination.) These crustal proportions were calculated assuming simple binary mixing between basalt and crust with 1 8 7 O s / 1 8 8 O s = 0.6 and [Os] = 50 ppt, the average Os composition of the upper continental crust (Esser and Turekian, 1993). These calculated proportions will probably change when Os measurements of locally derived crustal rocks are obtained. However the sense of the trend in Fig. 2 will not change. Thus He and Os apparently provide coherent information about crustal contamination, despite their markedly different geochemical behaviour. We currently have Os isotopic data for two Oligocene flood basalts from the Northwestern Ethiopian Plateau. These samples were chosen to represent the most primitive end-members of two basaltic suites identified on the basis of major and trace element characteristics (Pik, 1997): 1) A low-Ti suite similar to those frequently observed in other flood basalt provinces, often thought to display lithosphere-derived features; and 2) A high-Ti suite with OIB isotopic and trace element characteristics, notably high 3He/4He ratios suggestive of derivation from the deep mantle (Marry et al., 1996). (A third suite of intermediate character is currently being analysed.) The low-Ti sample, though relatively primitive (Mg# = 63.2), has a low Os concentration (13 ppt). Its initial lSTOs/188Os ratio (0.151) has