Pure path phase velocities of Love and Rayleigh waves across the Tibet Plateau are determined for the period range 30--90 s and for two opposite directions of propagation along the length of the Plateau. The method used is a two-event method, similar in principle to the classical two-station method, where the roles of stations and events are interchanged. The success of this method is due to a favorable geometry, whereby events located at each end of the plateau align to within 1¿ of a great circle path with one WWSSN station at teleseismic distance to the east and one to the west. Body wave modeling is used to constrain depth and source mechanism of the events in order to calculate the corresponding source phases for surface waves. Phase velocities are determined to within ¿0.02 km/s or better for Rayleigh waves and ¿0.007 km/s for Love waves. The constraints added by these data to the crust and upper mantle structure beneath Tibet are discussed in the light of previous studies. The best fitting model has a 65 km thick crust, with low average crustal velocities and an Sn velocity of 4.65 km/s. The low velocity zone is limited to the depth range 100--150 km. The interesting result of this study is that both love and Rayleigh wave data can be matched to within 1% in phase velocity by a unique isotropic model. This points to a similarity with a stable region upper mantle structure such as that of the Canadian Shield. Finally, additional Rayleigh wave phase velocity data are used to compare propagation on different paths in the vicinity of Tibet. The Tarim Basin, to the north of Tibet, stands out as having a distinct crustal structure, more similar to an average ''tectonic region'' structure. Moreover, evidence is presented for the existence of a very low velocity region at the western end of Tibet, which correlates well with a pronounced gravity low observed in that area.