The measurement of split PS-wave time delays and polarization directions is notoriously difficult in field data, partly because the signals are small and overprinted by competing mechanisms. This contribution describes a new processing method that suppresses many of the overprinted traveltime and amplitude anomalies, allowing depth-averaged split PS-wave time delays and polarization directions to be measured simply and precisely. These depth-averaged properties are then inverted using a simple forward model, allowing an earth model of split PS-wave time delays and polarization azimuths to be estimated without the need for layer stripping. In the field data used as an example, the processing and inversion methods are used to estimate split PS-wave time delays and polarization directions for ten layers spanning about 500 m depth from the seabed downward. Inversions using data-error covariances estimated from prestack data show model uncertainties less than 0.3 ms of time delay and 3° of polarization azimuth. However, it is clear that if the data-error covariances cannot be estimated from prestack data, due to low fold for example, model uncertainties would rise considerably. Repeating the inversions using data-error covariances of a postulated form leads to a range of maximum-likelihood models. When the data-error covariances cannot be estimated from prestack data, it seems reasonable to report precision levels implied by the spread of maximum-likelihood models, which in this case is up to 0.5 ms of time delay and 20° of polarization azimuth. The principal achievement of this processing and inversion scheme is to constrain a relatively large number of depth layers with similar levels of model uncertainty. The depth resolution available to this new method may have important implications for the development of tight-gas and shale-gas plays, in which variations of stress, strain, and fracture properties in discrete layers are important.