Accurate velocity information is a critical factor in the seismic migration process with more accurate velocities generally yielding better image quality. A study was undertaken to compare the effect of using velocity functions obtained from conventional stacking velocities with those derived from refraction modeling. A multichannel line from the southeastern Grand Banks (GSC line 85-2) was selected for this study due to the existence of a high quality coincident refraction survey for which a velocity model had been developed. Post-stack Finite Difference migration was applied to the multichannel data using the original stacking velocities in one case and the modeled refraction velocities in the other. The results highlight subtle but important differences in image quality. Specific examples show that complex structure associated with faulting is better resolved on sections migrated using refraction velocities, despite the relative coarseness of the spatial sampling of the velocity model. Contrary to expectations, migration using refraction velocities did a better job at relatively shallow depths (~1 s) than when stacking velocities were used. The seismic image still displays a number of artifacts including diffractions from dipping events and persistent water bottom multiples beneath the slope which are exacerbated by overmigration effects. Subsequent studies will focus on the creation of hybrid velocity models synthesized from both sources of velocity information and extending the work into the pre-stack domain.