To test driving force models for plate tectonics, the global intraplate stress fields predicted by various force systems are compared with the long-wavelength features of the observed stress field as determined by midplate earthquake mechanisms, in situ measurements, and stress-induced geologic structures. The calculated stresses are obtained by a finite difference solution to the equilibrium equations for thin elastic spherical shells in the membrane state of stress. Buoyancy forces at spreading centers and convergent plate boundaries and viscous drag at the base of the lithosphere are modeled as surface tractions applied to the shell. Drag is modeled as both a resistive and a driving force, and both symmetric and nonsymmetric forces at subduction zones are considered. The net driving push at spreading centers is found to be at least comparable in magnitude to other forces acting on the lithosphere and in particular is 0.7 to 1.5 times the net driving pull at convergence zones. Subducting lithosphere, which from seismic and thermal evidence has more potential energy available to drive plates than does a spreading center, thus converts relatively little of this energy to a net force acting on the surface plates. The drag coefficient at the base of the lithosphere may be greater by a factor of 3 to almost 10 beneath continents than beneath oceans without substantially affecting the fit between calculated and observed stress fields. Intraplate stresses calculated for models in which viscous drag at the base of the lithosphere acts in the direction of absolute plate velocity to drive plate motion are in much poorer agreement with observed stresses than are those calculated for models in which drag resists plate motions.