Free-air gravity anomalies across the southeast Greenland margin are investigated in conjunction with a well-constrained seismic velocity model to provide a constraint on subsurface density structure. This volcanic rifted margin is characterized by the presence of ~30-km-thick igneous crust, which correlates with a positive gravity high of ~60 mGal. A new systematic approach is adopted for gravity modeling, which consists of (1) full error propagation from the velocity model to predicted gravity anomalies through a posteriori model covariance represented by Monte Carlo ensembles, (2) the inversion of residual gravity anomalies for density variations within geological subdomains, and (3) the joint inversion of seismic travel times and gravity anomalies. A density model derived from the velocity model, using conventional conversion laws for the continental and oceanic crust, substantially underpredicts the observed gravity by ~70 mGal over the continental shelf. Neither errors in the velocity model nor the uncertainty of the chosen conversion laws are shown to be sufficient for such a large gravity misfit. A possible range of mantle contribution is first investigated by modeling various thermal evolution and depletion scenarios, which suggests that the maximum contribution is only ~20 mGal, assuming constant source mantle composition throughout continental rifting and subsequent seafloor spreading. If most of the residual gravity anomaly has a crustal origin, applying a conversion law with a denser upper crust in the continent-ocean transition zone seems to be the only plausible option to resolve this difficulty. Contrasting eruption environments for the transition zone crust (submarine) and the oceanic crust (submarine) probably result in different porosity structures, to which a velocity-density relationship is highly sensitive. The wire log and laboratory measurements of plateau basalts recovered from recent drilling legs on North Atlantic margins seem to support this explanation. An alternative explanation, which invokes a strong degree of source mantle heterogeneity, is also plausible on the basis of a recent geochemical study of the North Atlantic igneous province. ¿ 2001 American Geophysical Union