Select models of the main geomagnetic field and its secular variation are extrapolated to the base of an insulating mantle and used to estimate the adjacent fluid motion of a perfectly conducting outer core. The assumption of steady motion provides formally unique solutions and is tested along with that of no upwelling. The hypothesis of no upwelling is found to be substantially worse than that of steady motion. Although the actual motion is not thought to be steady, the large-scale secular variation at the top of the core can be adequately described by a large-scale, combined toroidal-poloidal circulation which is steady for intervals of at least a decade or two. The derive flows include a bulk westward drift but are complicated by superimposed jets, gyres, and surface divergence indicative of vigorous vertical motion at depth. The circulation pattern and key global properties including rms speed, upwelling, and westward drift are found to be fairly insensitive to variations in modeling parameters.