The rifting of continents and evolution of ocean basins is a fundamental component of plate tectonics, yet the process of continental break-up remains controversial. Plate driving forces have been estimated to be as much as an order of magnitude smaller than those required to rupture thick continental lithosphere 1,2 . However, Buck 1 has proposed that lithospheric heating by mantle upwelling and related magma production could promote lithospheric rupture at much lower stresses. Such models of mechanical versus magma-assisted extension can be tested, because they predict different temporal and spatial patterns of crustal and upper-mantle structure. Changes in plate deformation produce strain-enhanced crystal alignment and increased melt production within the upper mantle, both of which can cause seismic anisotropy 3 . The Northern Ethiopian Rift is an ideal place to test break-up models because it formed in cratonic lithosphere with minor far-field plate stresses 4,5 . Here we present evidence of seismic anisotropy in the upper mantle of this rift zone using observations of shear-wave splitting. Our observations, together with recent geological data, indicate a strong component of melt-induced anisotropy with only minor crustal stretching, supporting the magma-assisted rifting model in this area of initially cold, thick continental lithosphere.The data we analysed were collected as part of the EAGLE project (Ethiopian Afar Geophysical Lithospheric Experiment), an international multi-institutional experiment designed to investigate rifting processes in Ethiopia 6 . The Miocene-Recent Ethiopian Rift (Fig. 1) constitutes the northern part of the East African Rift system and forms one arm of a triple junction that formed on or near a mantle plume. Our study region is transitional between continental and incipient oceanic, with strain localized to ,20-km-wide zones of dyking, faulting and volcanism 6,7 . It is an ideal place to study magmatism and plate rupture, because up to 25% of the crust is extruded lava or intrusive magma 7,8 and mantle lithosphere is thin (,50 km) beneath the rift valley 9 . Seismic data were acquired in three phases of the EAGLE project 6 , two of which were designed to record passive seismicity. In phase I, 29 broad-band seismometers were deployed for 16 months with a nominal station spacing of 40 km and covering a 250 km £ 350 km region centred on the transitional part of the rift (Fig. 1). In phase II, a further 50 instruments were deployed for three months in a tighter array (nominal station spacing of 10 km) in the rift valley. Our study of mantle anisotropy is based on evidence of shear-wave splitting in the teleseismic phases SKS, SKKS and PKS recorded by these two arrays. With the longer duration array, 15 events produced usable splitting results, and with the shorter duration rift-valley array, three events produced usable results (list of events given in Supplementary Information).Shear-wave splitting analysis of the seismic phases SKS, SKKS and PKS is now a standard tool for studyi...