S U M M A R YThe processes involved in continental collisions remain contested, yet knowledge of these processes is crucial to improving our understanding of how some of the most dramatic features on the Earth have formed. As the largest and highest orogenic plateau on the Earth today, Tibet is an excellent natural laboratory for investigating collisional processes. To understand the development of the Tibetan Plateau, we need to understand the crustal structure beneath both Tibet and the Indian Plate. Building on previous work, we measure new group velocity dispersion curves using data from regional earthquakes (4424 paths) and ambient noise data (5696 paths), and use these to obtain new fundamental mode Rayleigh wave group velocity maps for periods from 5 to 70 s for a region including Tibet, Pakistan and India. The dense path coverage at the shortest periods, due to the inclusion of ambient noise measurements, allows features of up to 100 km scale to be resolved in some areas of the collision zone, providing one of the highest resolution models of the crust and uppermost mantle across this region. We invert the Rayleigh wave group velocity maps for shear wave velocity structure to 120 km depth and construct a 3-D velocity model for the crust and uppermost mantle of the Indo-Eurasian collision zone. We use this 3-D model to map the lateral variations in the crust and in the nature of the crust-mantle transition (Moho) across the Indo-Eurasian collision zone. The Moho occurs at lower shear velocities below northeastern Tibet than it does beneath western and southern Tibet and below India. The east-west difference across Tibet is particularly apparent in the elevated velocities observed west of 84• E at depths exceeding 90 km. This suggests that Indian lithosphere underlies the whole of the Plateau in the west, but possibly not in the east. At depths of 20-40 km our crustal model shows the existence of a pervasive mid-crustal low velocity layer (∼10% decrease in velocity, V s < 3.4 km s −1 ) throughout all of Tibet, as well as beneath the Pamirs, but not below India. The thickness of this layer, the lowest velocity in the layer and the degree of velocity reduction vary across the region. Combining our Rayleigh wave observations with previously published Love wave dispersion measurements, we find that the low velocity layer has a radial anisotropic signature with V sh > V sv . The characteristics of the low velocity layer are supportive of deformation occurring through ductile flow in the mid-crust.