A study of the results of the gravity and seismic surveys in Hudson Bay in 1965 has shown that the gravitational effect of a two-layer model based on the seismically determined depths has no correlation with the observed gravity anomalies. On the profile from Churchill to Povungnituk the gravity and seismic observations can be reconciled by postulating lateral variations of the acoustic compressional wave velocity within the crust. A crustal model has been calculated—using the same time-terms and the same mean crustal velocity—whose gravitational effect fits the observed gravity. The velocity varies from 6.15 to 6.56 km/s and the postulated depths are almost entirely within the confidence limits of the original model.In order to test the hypothesis, the postulated velocity variations have been compared with the lower refractor velocities of the shallow seismic survey, based on the assumption that the crustal velocities ought to be systematically higher than the crystalline surface velocities and that there may be a correlation between variations in crustal and surface velocities. The test is inconclusive because bottom refractor velocities are higher than crustal velocities in two areas where volcanic flows and high-velocity sediments may be present.The case of linearly related velocity (V) and density (ρ) variations has been analyzed and it is shown that the gravitational effect of the crust–mantle boundary undulations may be completely masked or even overbalanced by density changes in the crust if . The crust can be characterized by having dominant velocity variations (in which case the gravity anomaly reflects the undulations of the crust–mantle boundary) or dominant density variations (in which case the gravity anomaly inversely reflects the crust–mantle boundary undulations) depending on the relationship between average crustal density and average crustal velocity.