Seismic anisotropy is one of the most efficient geological and geodynamical tools for understanding the dynamics of the Earth. Upper mantle anisotropy is evident in seismic data sets for the last 30 years primarily from surface wave dispersion curves and body wave SKS data. We demonstrate in this paper that surface wave and body wave derived anisotropy can be explained by the same anisotropic parameters (L,Gc,Gs) in the simplest case of a horizontal fast symmetry axis. One application of this method is to display the SKS delay time and the corresponding azimuth, which can be derived from surface wave global (with a lateral resolution of around 2000 km) and regional (lateral resolution of 350 km) anisotropy tomography. A global-scale comparison is disappointing since there are only a few areas where both data types are correctly retrieved at the same spatial scale (same lateral resolution). The anisotropy is well resolved in oceanic basins from surface waves, whereas most measurements of SKS splitting are available below continents. However, there is a good agreement between surface wave and body wave anisotropy in regions where large-scale processes (primarily tectonic) are taking place, such as the western United States and in central Asia. In particular, we show that, at high frequency (>1 Hz), the observed SKS delay time and azimuth depend on the order in which S waves propagate through the layers. A comparison between synthetic and observed SKS phases in central Asia highlights the varying sensitivity of SKS waves at depth. We can obtain a good correlation between synthetic and observed SKS by considering anisotropy only between 80 and 200 km depth in the tomographic model, rather than in the entire upper mantle. However, it is more difficult to explain the discrepancy between synthetic and observed SKS under the eastern United States, where a large-scale coherent SKS anisotropy is present. Such results call for more sophisticated SKS waveform modeling and for enhanced lateral resolution in anisotropic tomographic models. ¿ 2000 American Geophysical Union