Problems associated with the application of topographic corrections to deep-ocean marine seismic data are described. Synthetic travel time data are used to show that corrections for the delay time introduced by the water column are inferior to corrections for the true water path. Water path corrections place the shots at the ray entry points on the seafloor, require no assumptions about the structure under investigation, and maintain the correct distribution of data with range. The water path correction can be most accurately applied during the stacking operation that transposes the observed data into intercept time-slowness space. There is, then, no requirement to make determinations of slowness, and for experiments involving seafloor receiver and surface shots the data may be transposed to the seafloor with complete objectivity. Water path corrected stacks from different locations may then be compared in the sure knowledge that any observed differences are due to real structural changes and not to topographic correction errors or to invalid assumptions of lateral homogeneity.