We collected shallow reflection data in southwestern Montana, USA, across a 5.4-m-high tectonic scarp. The goal was to image the normal fault associated with the scarp, observed in an adjacent trench. Processing of the data was challenging because the height of the scarp was comparable to the depths of the reflectors of interest. To find out how to proceed, we processed synthetic data generated using velocity models derived in part from actual shot gathers. The actual data are dominated by large-amplitude low-frequency surface waves, but clear high-frequency reflections are seen in the more distant geophones. Common-offset gathers for the raw and high-pass filtered data reveal sharp discontinuities in arrival times and a strong decrease in amplitudes, respectively, under the scarp. These changes in the wavefield are indicative of lateral variations in elastic properties and are consistent with the presence of a fault zone seen in the trench. The actual data were stacked after the surface waves were removed with a narrow f-k filter. Severe muting was applied to isolate the reflections seen in the high-pass filtered data. The stacked data reveal a clear and fairly continuous horizontal reflector on the downthrown side of the fault and more disrupted reflectors on the upthrown side, with truncated reflections and changes in amplitude roughly across the projection of the fault mapped in the trench. These observations are consistent with faulting and would be difficult to explain if the scarp were an erosional feature.