A month of digital data from two three component seismograph stations near Wellington, New Zealand, was analysed as part of a feasibility study for a major project to investigate shear-wave splitting. Although the total number of earthquakes studied was small (14), some suggestive results were obtained. Almost all events recorded within the shear wave window showed a phase reversal of the horizontal components after one or two shear wave cycles, suggesting that there are actually two shear-wave arrivals. The measured polarization of the first shear wave arrivals was N (21+/-11) E. This polarization alignment cannot be explained by focal mechanisms, and it is unlikely to be due to topography because of the station-to-station correlation. The present evidence suggests the most likely cause is crustal anisotropy due to the geological structure at shallow depth, rather than stress aligned micro-cracks. ¿ American Geophysical Union 1990