In this study, the SV autocorrelation (SVA) technique of Dasgupta and Nowack (2006) is tested using three-component P-wave data from selected INDEPTH II and CDSN stations in southern Tibet. The SVA technique involves constructing an estimate of the source and distant Earth wavelet from the autocorrelation of the SV component and using this to deconvolve the data. The deconvolved vertical components are then used to infer the crustal P-velocity structure. Initial models for the inversion of the deconvolved vertical components were based on the S-velocity results obtained from the inversions of radial receiver functions by Mitra et al. (2005). The Moho depths obtained from inverting the vertical components deepen from the south to the north in the study area and are comparable to the depths obtained by Mitra et al. (2005) to the north and similar but somewhat shallower to the south. The crustal VP/VS ratios are obtained from the inverted VS models from Mitra et al. (2005) and the inferred VP speeds obtained from this study. Except for station BB18 with a somewhat higher VP/VS ratio, the other stations to the north in the Tethyan Himalayas and the southern Lhasa terrane have VP/VS ratios between 1.75 and 1.80 with a mean of 1.77 (a Poisson’s ratio of 0.265). The range of values could result from noise in the data as well as from lateral heterogeneity in the region with different piercing points at depth for the Ps and PpPp phases.