Seismic coherence is routinely used to delineate geologic features that might otherwise be overlooked on conventional seismic amplitude volumes. In general, one wishes to interpret the most broadband data possible. However, because of the thickness tuning effects, certain spectral components often better illuminate a given feature with higher signal-to-noise ratio than others. Clear images of channels and other stratigraphic features that may be buried in the broad-band data may "light up" at certain spectral components. For the same, coherence attributes computed from spectral voice components (equivalent to a filter bank) also often provide sharper images, with the "best" component being a function of tuning thickness and reflector alignment across faults. While one can co-render three coherence images using RGB blending, display of the information contained in more than three volumes in a single image is difficult. We address this problem by summing a suite of structure-oriented covariance matrices computed from spectral voices resulting in a "multi-spectral" coherence algorithm. We demonstrate the value of multi-spectral coherence by comparing it to both RGB blended volumes and coherence computed from spectrally balanced, broad-band seismic amplitude volume from a .megamerge survey acquired over the Red Fork Formation of the Anadarko Basin, Oklahoma. The multi-spectral coherence images provide better images of channel incisement and are less noisy than those computed from the broadband data. Multi-spectral coherence also provides several advantages over RGB blended volumes: first, one can combine the information content from more than three spectral voices; second, only one volume needs to be loaded into the workstation; and third, the resulting gray-scale images can be co-rendered with other attributes of interest, for example, petrophysics parameters, plotted against a polychromatic color bar.