We have obtained constraints on mechanical crust–mantle coupling for Tibet and Yunnan/Indo China, by comparing the observed surface deformation field inferred from GPS and Quaternary fault slip rate data, with the mantle deformation field inferred from several SKS shear wave splitting data sets. We first determined whether the anisotropy is dominantly asthenospheric or lithospheric by testing simple models of both types against the observed values of the fast polarization direction, ϕ, which is assumed to be parallel to the horizontal projection of the direction of maximum shear. For asthenospheric flow, we solved for a best-fitting uniform sub-asthenospheric velocity model for Eurasia. The fit, however, was not satisfactory (RMS misfit Δ ϕ = 20°). Solving for separate uniform flow fields in each region improves the fit (Δ ϕ = 15° for Tibet, Δ ϕ = 11° for Yunnan), although the resulting flow fields are inconsistent with several geophysical and geological constraints and thus considered unlikely. We then considered lithospheric models. For Tibet, vertically coherent deformation (i.e., maximum shear direction from surface deformation is parallel to ϕ) yields an improved match (Δ ϕ = 11°) for left-lateral shear. Both the goodness of fit and the dominance of left-lateral surface faulting in Tibet, argue for a lithospheric source of anisotropy. The misfit for Yunnan is large for either right- (Δ ϕ = 53°) or left-lateral (Δ ϕ = 49°) shear, which argues for complete crust–mantle decoupling in Yunnan. We show that the fast polarization directions throughout both the Tibet and Yunnan region can be fit by a single lithospheric dynamic model in which there is strong coupling between crust and mantle beneath Tibet, but a complete decoupling between crust and mantle beneath Yunnan crust. This dynamic model predicts left-lateral maximum shear directions within the mantle that align with fast polarization directions in both regions (Δ ϕ = 9°). These maximum shear directions within the mantle align with the left-lateral maximum shear directions in the crustal deformation field in Tibet, but are not in Yunnan. Our results have the following implications. First, the coherence between crust and mantle deformation in Tibet implies strong crust–mantle mechanical coupling, since this is the only way that crustal buoyancy forces, required to account for the surface deformation field, can be transmitted into the mantle. This behavior is consistent with a uniform-strength lithosphere or strong crust, but not with a substantially weaker crust. For example, it is inconsistent with the popular “jelly-sandwich” rheology, and thus precludes behavior such as large-scale lower crustal flow in Tibet. The crust–mantle coherence is also incompatible with mantle delamination. Second, crust–mantle decoupling within the Yunnan lithosphere argues that mantle deformation there is controlled only by boundary conditions; crustal buoyancy forces are not transmitted into the mantle beneath Yunnan. Moreover, the dynamic model for the mantle shows that the Yunnan crust is moving south with respect to the mantle at rates as high as ∼30 mm/yr. Third, there is a fundamental rheological lithospheric transition between Tibet and Yunnan that may provide a key to understanding this significant orogen.