Experiments have been conducted to demonstrate the accuracy and precision of moisture content estimates derived from cross-borehole ground penetrating radar (XBGPR) measurements made within the vadose zone. Both numerical simulations and field data demonstrate that although a certain amount of image smearing occurs under ideal conditions the general trends in the spatial variation of the moisture content can be estimated by a simple empirical transformation from images of electromagnetic (EM) wave velocity. The field results are verified by comparing the radar-derived images of volumetric moisture content to neutron log derived values. When an appropriate site-specific conversion from radar wave velocity to moisture content is applied, a root mean square (RMS) error of 2.0--3.1% volumetric moisture content exists between the two sets. Further comparison of the two different data sets along with analysis of plots of the ray density through each cell indicates that regions of high moisture content are better resolved than regions of low moisture and that most of the discrepancy between radar-derived and neutron-derived moisture contents occurs in regions of high moisture content. Better spatial resolution can be provided if dense station spacing is used. However, the amount of extra time required to acquire the extra data may limit the usefulness of the method. Repeatability measurements made with five data sets demonstrate that the precision error of the data acquisition system employed averages about 0.54 ns, which translates to about a 0.5% error in moisture content estimation.