Description: A 550-km-long transect across the Ninetyeast Ridge, a major Indian ocean hotspot trail, provided seismic refraction and wide-angle reflection data recorded on 60 ocean bottom instruments. About 24 000 crustal and 15 000 upper mantle arrivals have been picked and used to derive an image of the hotspot track. Two approaches have been chosen: (i) a first-arrival tomographic inversion yielding crustal properties; and (ii) forward modelling of mantle phases revealing the structure at the crust–mantle boundary region and of the uppermost mantle. Away from the volcanic edifice, seismic recordings show the typical phases from oceanic crust, that is, two crustal refraction branches (Pg), a wide-angle reflection from the crust–mantle boundary (PmP) and a wave group turning within the upper mantle (Pn). Approaching the edifice, three additional phases have been detected. We interpret these arrivals as a wide-angle reflection from the base of material trapped under the pre-hotspot crust (Pm2P) and as a wide-angle reflection (PnP) and its associated refraction branch (PN) from a layered upper mantle. The resulting models indicate normal oceanic crust to the west and east of the edifice. Crustal thickness averages 6.5–7 km. Wide-angle reflections from both the pre-hotspot and the post-hotspot crust–mantle boundary suggest that the crust under the ridge has been bent downwards by loading the lithosphere, and hotspot volcanism has underplated the pre-existing crust with material characterized by seismic velocities intermediate between those of mafic lower crustal and ultramafic upper mantle rocks (7.5–7.6 km s−1). In total, the crust is up to ≈ 24 km thick. The ratio between the volume of subcrustal plutonism forming the underplate and extrusive and intrusive volcanism forming the edifice is about 0.7. An important observation is that underplating continued to the east under the Wharton Basin. During the shield-building phase, however, Ninetyeast Ridge was located adjacent to the Broken Ridge and was subsequently pulled apart along a transform fault boundary. Therefore, underplating eastwards of the fracture zone separating the edifice from the Wharton Basin suggests that prolonged crustal growth by subcrustal plutonism occurred over millions of years after the major shield-building stage. This fact, however, requires mantle flow along the fossil hotspot trail. The occurrence of PnP and PN arrivals is probably associated with a layered and anisotropic upper mantle due to the preferential alignment of olivine crystals and may have formed by rising plume material which spread away under the base of the lithosphere.