Where they can be correlated with geological exposures and trenches, 3-D ground-penetrating radar (GPR) data can contribute critical subsurface information to paleoseismic investigations. Because active faults are typically characterized by complicated near-surface structures that vary with the styles of faulting and the types of rock that are ruptured, GPR data can be difficult to interpret. We have acquired 3-D GPR data sets across three active fault zones within New Zealand that have different deformation styles: the strike-slip Wellington fault zone, reverse faults of the Ostler fault zone, and normal faults of the Maleme fault zone. To improve our interpretation of the processed GPR volumes, we employed two suites of geometric attributes. The first suite was computed using a coherence-based algorithm. It provided estimates of the coherency, azimuth, and dip of reflections. The second suite quantified the volumetric textures of reflections, which allowed different reflection facies to be defined objectively. We have demonstrated how some attributes were more successful at visualizing certain structural or depositional characteristics than others. For example, the coherency attribute was an excellent tool for highlighting normal faults within volcanic deposits of the Maleme fault zone, whereas the texture-based attributes were most useful for discriminating between the gravel and metasediment units juxtaposed by the Wellington fault zone. Our GPR data sets and associated attribute volumes showed details of near-surface fault geometry that were not obvious from surface mapping, and they revealed evidence of off-fault deformation, gravitational collapse, and topple structures.