The sense of shear traction which the deeper mantle exerts on the North America plate is controversial. To test this, we compute laterally-varying thin-shell models. Fault elements are used to outline the plate, so the velocity of North America is not fixed. The basal boundary condition is set in one of three ways: (a) for resistive drag, we assume that the lower mantle is static with respect to Africa; (b) we test models with no basal traction; (c) for active drag, we assume that the lower mantle moves as a rigid plate but 10% faster than North America. Each model is scored by comparison with sea-floor spreading rates, geodetic velocities, stress directions, and NUVEL-1A. Only models with active drag are successful. While these results do not determine the exact azimuth or pattern of basal drag on the North America plate, they establish the modal sense as active.