The relative northward motion of the Indian continent, after collision with Eurasia, caused deformation that resulted in high elevations and crustal thickening throughout much of Asia. One of the most spectacular features from this collision is the Tibetan Plateau. The Tibetan Plateau extends from the south side of the Kunlun Shan to the north side of the Himalaya Range and consists of three micro-continental fragments that accreted onto the southern margin of Eurasia after the breakup of Gondwana. Continental accretion ended with the collision of the Indian subcontinent approximately 40 my that produced nearly 2000 km of shortening as well as the high elevations (5000 m) and thickened crust (60-70 km). Most models attribute the characteristics of the Tibetan Plateau to underthrusting of the Indian lithosphere beneath the southern margin of Eurasia, distributed shortening and thickening of the Eurasian lithosphere and lateral extrusion of crustal blocks. Accurate determination of the upper mantle velocity structure of the Tibetan Plateau is thought to be a key to understanding the tectonic evolution of this important region. Compared to regions throughout the world, relatively little is known about the deep structure beneath the Tibetan Plateau. Consequently, models describing the tectonic evolution of the plateau are not well constrained. Previous studies of Tibetan earthquakes have placed first-order constraints on upper mantle velocities, but have to rely on data recorded almost exclusively at stations outside of the plateau. The opportunity to make direct seismic observations within the plateau offers the possibility to vastly improve our understanding of the deep structure of this region. Results are presented here are from the analysis of teleseismic and regional phases recorded within the Tibetan Plateau. The most significant observation is that lateral variations exist within the upper mantle despite the relatively uniform topography of the plateau. We observe large values of shear-wave splitting, slow Pn velocities and inefficient Sn propagation in the northern plateau relative to the south. From these observations we infer that the lithospheric mantle beneath the northern plateau has elevated temperatures relative to the south and is partially melted. We also observe that deformation structure within the uppermost mantle is coherent with structural trends at the surface. Lateral variations in the mantle provide constraints on models describing the tectonic evolution of the collision between the Indian and Eurasian plates. A combination of several previously proposed tectonic models provides a reasonable explanation to our observations. Specifically, we envision an Indian plate that is underthrust at a shallow angle toward the north with significantly steeper, southward underthrusting of the Eurasian lithospheric mantle. The lithospheric mantle dynamics, hypothesized here are not original ideas. Similar tectonic models have been proposed to describe the interaction between the Eurasian and Indian plates, in the Pamirs and Karakoram. We merely suggest that plate configuration beneath the Pamirs and Karakoram continues eastward, into the broader Tibetan Plateau.