We apply a two-station method of velocity determination to Pn arrival times reported in the International Seismological Centre Bulletin during the period 1964--1993. The azimuthal variation of these velocities provides evidence of Pn anisotropy as large as 9.6%. Anisotropic regions are identified by examination of the azimuthal variation of the Pn velocities within circular caps of radius 1.5¿ and 3¿, spaced at 1¿ intervals across the Earth. This procedure leads to 255 geographically distributed estimates of anisotropy. The length scale over which the anisotropy varies indicates that uppermost mantle anisotropy beneath continents is controlled by regional orogenic strains and is inconsistent with a general interpretation in terms of plate-scale mantle flow. In addition, the correlation between Pn anisotropy and surficial tectonic features indicates that the anisotropy is primarily the result of the most recent deformation event. We compare our Pn anisotropy results (fast directions and magnitudes) with those of SKS splitting studies to attempt to constrain the depth extent of the anisotropy. This comparison suggests that in some areas, where the fast directions agree, SKS splitting observations may arise from a layer beginning just below the Moho, whereas in others it must have a deeper origin. In these areas, the relative magnitudes of the measured anisotropy are consistent with a thin (100--150 km) lithospheric layer of anisotropy. ¿ 1999 American Geophysical Union