Channel-form morphologies (channel or valley) proliferate in the rock record and are the target of many exploration and development wells. Their geometry and the nature of their relationships with underlying, laterally adjacent, and overlying strata are often complex and difficult to deduce with 1D, 2D, and even 3D data due to their limited spatial extent and thickness. Threedimensional seismic has added a new dimension to the exploration of and for channelform deposits, allowing many attributes to be extracted from these data in an attempt to reveal details of channelized deposits. Identifying which among this multitude of attributes are actually useful for mapping channel-form nature (planform, depth, edges, internal architecture, lithology, etc.) is a long-standing challenge for petroleum workers and a future challenge for groundwater workers. A high-quality, large (3,154-km2) 3D seismic data volume collected in a highly channelized stratigraphic series found in the upper 500 ms of the Sunda Shelf in offshore Indonesia was used to map incised valley development on this shelf. Defined on a seismic grid at 25-m spacing, ten valley bases were then exported to ArcGIS, which was used to define each valley’s thalweg. Valleybase elevation and several seismic attributes were then extracted along each thalweg. Resulting correlations and relationships could be used to infer thickness variability and reservoir characteristics of channel forms from 2D planform analysis and profiling. Two seismic attributes, coherency and 30-Hz component amplitude, are somewhat correlated (|r|>0.23) with along-valley thicknesses collectively across the ten valleys. Instantaneous frequency, reflection strength, and average absolute amplitude exhibit negligible correlation with thickness (|r|<0.03). Spectral decomposition is the most intuitive method for attribute-to-thickness correlation, although a general relationship between spectral component amplitude and thickness of imaged features has not yet been determined. Thickness of Valley 1, the thickest valley analyzed, is best-correlated (r = 0.62) with maximum 10-Hz component amplitudes within the lower interval of the valley fill.