The notion that plate motions result from forces associated with plate boundaries implies that average plate velocities in a mantle (hot spot) frame of reference are systematically related to the nature and geometry of plate boundaries, and to the proportion of continental lithosphere included in the plates. Several simple, linear models relating plate velocity &ngr; to net ridge push (RP), slab pull (SP), trench suction (TS) and continental drag (CD) have been evaluated by multiple linear regression. Ridge push, slab pull and continental drag are found to be vitally important. Trench suction may be a real force, but because effective trench lengths are small and only three of the twelve plates considered are affected, it is not a statistically significant contributor to plate motions. The preferred relation, &ngr;(cm/yr) = (2.7¿4)+(5.2¿1.9)RP+(13.2¿1.6SP -(4.7¿.6)CD is a remarkably good predictor of average plate velocities, with an rms error of 0.50 cm/yr, and a correlation coefficient of 0.97. The implied balance between these net forces is generally consistent with the results of force-balance models. The constant term includes all forces not explicitly considered, and suggests that the combined effects of such forces are about equal for all plates. Finally, because the parameters used for any particular plate are independent of all other plates, the success of this simple model strongly suggests that the forces which drive plate motions are intimately related to the plates themselves.