Mantle upwelling is essential to the generation of new oceanic crust at mid-ocean ridges, and it is generally assumed that such upwelling is symmetric beneath active ridges. Here, however, we use seismic imaging to show that the isotropic and anisotropic structure of the mantle is rotated beneath the East Pacific Rise. The isotropic structure defines the pattern of magma delivery from the mantle to the crust. We find that the segmentation of the rise crest between transform faults correlates well with the distribution of mantle melt. The azimuth of seismic anisotropy constrains the direction of mantle flow, which is rotated nearly 10u anticlockwise from the plate-spreading direction. The mismatch between the locus of mantle melt delivery and the morphologic ridge axis results in systematic differences between areas of on-axis and off-axis melt supply. We conclude that the skew of asthenospheric upwelling and transport governs segmentation of the East Pacific Rise and variations in the intensity of ridge crest processes. The origin of segmentation of oceanic spreading centres is controversial. According to one point of view, along-axis differences in ridge crest processes result directly from three-dimensional mantle upwelling 1-4. Sites of vigorous volcanic and hydrothermal activity are thus thought to overlie regions of greater magma supply. Limited knowledge of mantle structure, however, has given rise to diverging opinions on the scale of three-dimensional upwellings 2-6. Alternatively, segmentation of ridge crest processes may be regulated by the tectonic rifting of young lithosphere 7 , and thus not directly linked to the form of mantle upwelling. In this case mantle flow could be either two-dimensional 8 or three-dimensional and less obvious because of efficient along-axis transport of magma by viscous flow 9. Along the global ridge system, the fast-spreading East Pacific Rise (EPR) between the Siqueiros and Clipperton transforms (Fig. 1) currently offers our best opportunity for understanding the relations between mantle upwelling and ridge crest processes. There are several reasons for this. This section of the EPR encompasses a full spectrum of rise axis discontinuities, including: two large-offset transform faults; large, long-lived (9u 039 N) and small, short-lived (9u 379 N) overlapping spreading centres (OSCs) 10-12 ; and smaller-scale mor-phologic 3 , petrologic 5,12 and seismic 6,13,14 discontinuities that are typical of fast-spreading ridge segments. Accompanying these axial discontinuities are well known along-axis variations in seafloor depth, axial high morphology 15 , crustal structure and thickness 16-19 , lava chemistry 5,12,20,21 , and seafloor hydrothermal 22,23 and biological activity 22. These characteristics of the EPR, including the origin of ridge crest segmentation, have been hypothesized to result from the supply of magma from the mantle. We conducted the UNDERSHOOT experiment (our data-gathering cruise) to seismically image the crustal and mantle structure between the Clipperton and Siqueiros transforms to determine the pattern of magma delivery from the mantle to the crust. Here we present the first images of mantle structure beneath an entire ridge segment bounded by long-lived tectonic discontinuities (Fig. 1). Good image resolution allows direct comparison between the scales of segmenta-tion observed along this section of the EPR with the physical structure of the topmost mantle. Our results allow conclusions to be drawn about the driving and controlling processes for segmentation of fast-spreading ridges. Experiment geometry and tomographic imaging The distribution of seismic receivers and sources used to image crustal and mantle structure is shown in Fig. 1. The experiment constrains the structure of the uppermost mantle within 4 km of the Mohorovičić discontinuity and within an area extending 15 km to either side of the rise axis and 230 km along the spreading centre. A three-dimensional model of off-axis crustal structure and thickness is used to analyse the mantle refraction data (see Supplementary Information). The P n data (from the wave refracted below the Moho) provide good spatial sampling of mantle structure throughout the image volume (Fig. 2a). P n travel-time residuals plotted by azimuth reveal a cos2H pattern (Fig. 2b), a signal indicative of azimuthal seismic anisotropy. The azimuth of anisotropy (that is, the fast direction for P n propagation) is N73uE 6 1u (see Supplementary Information). Tomographic inversions, discussed below, confirm this result. The azimuth of anisotropy is rotated 9u anticlockwise with respect to the predicted spreading direction 24 (N82uE). Plotted by rise crossing point, P n travel-time residuals show evidence for anomalously low and variable upper-mantle velocities (Fig. 2c). The average isotropic velocity that best fits the P n data (7.6 km s 21) is less than typical upper-mantle velocities, whereas delays are greater towards the centre of the transform-bounded segment and less within 20 km of the transforms. Tomographic inversion (see Supplementary Information) of P n travel-time data reveals a mantle low-velocity zone (MLVZ) that is segmented on a scale comparable to tectonic offsets of the EPR (Fig. 1). The MLVZ decreases in amplitude towards each transform, in agreement with the decrease in mean P n delays (Fig. 2c). Between transforms, the MLVZ follows two en echelon trends that are orthogonal to the azimuth of seismic anisotropy (Fig. 1b, green lines). The en echelon trends are offset in a right lateral sense and rotated