Since its introduction two decades ago, coherence has been widely used to map structural and stratigraphic discontinuities such as faults, cracks, karst collapse features, channels, stratigraphic edges, and unconformities. With the intent to map azimuthal variations of horizontal stress as well as to improve the signal-to-noise ratio of unconventional resource plays, wide-/full-azimuth seismic data acquisition has become common. Migrating seismic traces into different azimuthal bins costs no more than migrating them into one bin. If the velocity anisotropy is not taken into account by the migration algorithm, subtle discontinuities and some major faults may exhibit lateral shifts, resulting in a smeared image after stacking. Based on these two issues, we evaluate a new way to compute the coherence for azimuthally limited data volumes. Like multispectral coherence, we modify the covariance matrix to be the sum of the covariance matrices, each of which belongs to an azimuthally limited volume, and then we use the summed covariance matrix to compute the coherent energy. We validate the effectiveness of our multiazimuth coherence by applying it to two seismic surveys acquired over the Fort Worth Basin, Texas. Not surprisingly, multiazimuth coherence exhibits less incoherent noise than coherence computed from azimuthally limited amplitude volumes. If the data have been migrated using an azimuthally variable velocity, multiazimuth coherence exhibits higher lateral resolution than that computed from the stacked data. In contrast, if the data have not been migrated using an appropriate azimuthally variable velocity model, the misalignment of each image results in a blurring of the multiazimuth coherence and the coherence computed from the stacked data. This suggests that our method may serve as a future tool for azimuthal velocity analysis.