The crustal and upper mantle azimuthal anisotropy of the Tibetan Plateau and adjacent areas was studied by Rayleigh wave tomography. We collected sufficient broadband digital seismograms traversing the Tibetan Plateau and adjacent areas from available stations, including especially some data from the temporary stations newly deployed in Yunnan, eastern Tibet, and western Sichuan. They made an adequate path coverage in most regions to achieve a reasonable resolution for the inversion. The model resolution tests show that the anisotropic features of scope greater than 400 km and strength greater than 2% are reliable. The azimuthal anisotropy pattern inside the Tibetan Plateau was similar to the characteristic of tectonic partition. The crustal anisotropy strength is greater than 2% in most regions of East Tibet, and the anisotropy shows clockwise rotation surrounding the eastern Himalayan syntaxis. Vertically, the anisotropy direction indicates a coherent pattern within the upper crust, lower crust, and lithosphere mantle of the Tibetan Plateau, which also is consistent with GPS velocity field and SKS fast polarization directions. The result supports that the crust-mantle deformation beneath the Tibetan Plateau is vertically coherent. The anisotropy strength of crust and lithospheric upper mantle in Yunnan outside the Tibetan Plateau is lower than 2%, so SKS splitting from core-mantle boundary to station should largely be attributed to the anisotropy of asthenosphere.Tibetan Plateau, azimuthal anisotropy, Rayleigh wave tomography, vertically coherent deformationThe formation of Tibetan Plateau is a result of collision between the Indian plate and Eurasian plate. Since the 1970s, Earth scientists have put forward several plate collision models of the Indian and Eurasian plate in accordance with different uplift and deformation mechanisms of Tibetan Plateau, for example, the kinematic model of extrusion or escape [1] and the kinetic model of continuous deformation [2] . Deep geophysical observation provides important information to evaluate the correctness of these models. To utilize broadband seismic data from Tibetan Plateau and surrounding areas to verify the collision model will definitely deepen the understanding of plate collision course.