Shear‐wave [Formula: see text]-wave) polarization azimuths, although consistent over large depth intervals, changed abruptly and by large amount of various depths in nine-component vertical seismic profiling (VSP) data from the Cymric and Railroad Gap oil fields of the southwest San Joaquin basin. A simple layer‐stripping technique made it possible to follow the polarization changes and determine the [Formula: see text]-wave birefringence over successive depth intervals. Because the birefringence and polarization azimuth are related to in‐situ stresses and fracture, information from such analysis could be important for reservoir development. Near offset VSP data from Cymrix indicated that the subsurface could be appproximated roughly as two anisotropic layers. The upper layer, from the surface to 800 ft (240 m), had vertical [Formula: see text]-wave birefringence as large was about 6 percent down to 1300 ft (400 m). In the upper layer the polarization azimuth of the fast [Formula: see text]-wave was N 60°E, while in the lower layer it was about N 10°E. Refinement of the layer stripping showed that neither layer was anisotropically homogenous, and both could be subdivided into thinner layers. Near offset [Formula: see text]-wave VSP data from the Railroad Gap well also show high birefringence near the surface and less birefringence deeper. In the uppermost layer, which extends down to 1300 ft (400 m), the [Formula: see text]-wave birefringence was 9 percent, and the lag between the fast and slow [Formula: see text]-waves exceeded 60 ms at the bottom of the layer. Seven layers in all were needed to accommodate [Formula: see text]-wave polarization changes. The most reliable azimuth angle determination as judged from the data consistency were those of the uppermost layer, at N 46°E, and those from depths 2900–3700 ft (880–1130 m) and 3900–5300 ft (1190–1610 m), at N 16°E and N 15°W, respectively. Over those intervals the scatter of calculated azimuths about the mean was typically less than 4 degrees. The largest birefringence at both locations occurred in the same formation, the Pliocene Tulare sands and Pebble Conglomerate. In those formations the azimuth of the fast [Formula: see text]-wave polarization was roughly orthogonal to the southwest. In the deeper Antelope shale, [Formula: see text]-wave polarization directions in both areas were close to 45 degrees from the fault. Confidence in the layer stripping procedure was bolstered by major improvement in data quality that resulted from stripping. Before stripping, wavelets of the two [Formula: see text]-waves sometimes had very different waveforms, and it was often impossible to come close to diagonalizing the 2 × 2 S‐wave data matrix by rotating sources and receivers by the same angle. After stripping, wavelets were more similar in shape, and the S‐wave matrix was more nearly diagonalizable by rotating with a single angle.