We applied time-domain seismic diffraction imaging to a 3D data set from the Piceance Creek Field, Piceance Basin, northwest Colorado. The work was motivated by the need for insight into natural fracture distribution, thought to influence production. We used a novel chain of two previously developed processing steps to separate diffractions from the recorded wavefield — One step is applied to the conventional stack volume, and the other was applied to migrated dip-angle gathers. The diffractions were then imaged independently for interpretation. Comparison of seismic attributes, commonly used for fracture characterization, found that the resulting diffraction image had lateral resolution comparable to or greater than the discontinuity-type attributes and provided information complementary to azimuthal anisotropy measurements. The diffraction image from Piceance Creek had advantages over attributes in interpretation confidence because diffractions were a direct seismic response to subsurface features of intermediate size. Although these features were larger than the fractures thought to influence production, knowledge of intermediate-scale features can improve fracture prediction in the context of geologic scaling relationships or rock physics models. Qualitative interpretation of the diffraction amplitudes distinguished edge-type and line-type diffractions, indicative of fault versus channel-fill features, respectively. Even the largest faults at Piceance Creek only generated diffractions where contrasting lithologies were juxtaposed. Where there was lateral contrast, diffractions appeared to delineate small faults and channels with vertical resolution limited to the same order as the conventional seismic image.