The upper mantle structure of Swedish part of Baltic Shield with its isotropic and anisotropic seismic velocity characteristics is investigated using telesesismic body waves (i.e. P waves and shear waves) recorded by the Swedish National Seismological Network (SNSN). Nonlinear high-resolution P and SV and SH wave isotropic tomographic inversions reveal velocity perturbations of ± 3 % down to at least 470 km below the network. Separate SV and SV models indicate several consistent major features, many of which are also consistent with P-wave results. A direct cell by cell comparison of SH and SV models reveals velocity differences of up to 4%. Numerical tests show that differences in the two S-wave models can only be partially caused by noise and limited resolution, and some features are attributed to the effect of large scale anisotropy. Shear-wave splitting and P-travel time residual analyses also detect anisotropic mantle structure. Distinct back-azimuth dependence of SKS splitting excludes single-layer anisotropy models with horizontal symmetry axes for the whole region. Joint inversion using both the P and S data reveals 3D self-consistent anisotropic models with well-defined mantle lithospheric domains. These domains of differently oriented anisotropy most probably retain fossil fabric since the domains' origin, supporting the idea of the existence of an early form of plate tectonics during formation of continental cratons already in the Archean. The possible disturbing effects of anisotropy on seismic tomography studies are investigated, and found to be potentially significant. P-wave arrival times were adjusted based on the estimates of mantle anisotropy, and re-inverted. The general pattern of the velocity-perturbation images was similar but changed significantly in some places, including the disappearance of a slab-like structure identified in the inversion with the original data. Thus the analysis demonstrates that anisotropy of quite plausible magnitude can have a significant effect on the tomographic images, and should not be ignored. If, as we believe, our estimates of anisotropy are reasonably correct, then the model based on the adjusted data should give a more robust and correct image of the mantle structure.