Until recently, the midmantle region of the Earth (the transition zone and the uppermost lower mantle) has been considered isotropic. Recent work by Wookey et al. <2002> presented evidence of anisotropy in the midmantle region near the Tonga-Kermadec subduction zone, inferred from shear wave splitting in teleseismic S phases at Australian seismic stations. This data set is revisited to examine the possibility of contamination by shear-coupled P waves. Using reflectivity modeling, we explore the effects of these phases on the measurement of shear wave splitting. Contamination of such measurements is demonstrated for short epicentral distances (Δ ~ 30¿). Wave field decomposition is a simple technique which can be used to separate these phases from the main S wave arrival. The Tonga-Australia data set is reprocessed after wave field decomposition. Shear wave splitting ranging between 0.7 and 6.2 s is observed, a range comparable to that observed by Wookey et al. <2002>. As before, the polarization of the fast shear wave is observed to be predominantly horizontal. However, the magnitudes of splitting for several events at short epicentral distances are significantly reduced, suggesting some influence of shear-coupled P waves in the original analysis. The results are compared with possible models of mantle anisotropy, and the results can be best explained by significant anisotropy in the midmantle region. The best candidate location for this anisotropy is in the uppermost lower mantle, and scenarios involving the alignment of lower mantle or subducted materials by regional dynamic processes are suggested to account for this.