A conceptual shift is overdue in geodynamics. Popular models that present plate tectonics as being driven by bottom-heated whole-mantle convection, with or without plumes, are based on obsolete assumptions, are contradicted by much evidence, and fail to account for observed plate interactions. Subduction-hinge rollback is the key to viable mechanisms. The Pacific spreads rapidly yet shrinks by rollback, whereas the subduction-free Atlantic widens by slow mid-ocean spreading. These and other first-order features of global tectonics cannot be explained by conventional models. The behavior of arcs and the common presence of forearc basins on the uncrumpled thin leading edges of advancing arcs and continents are among features indicating that subduction provides the primary drive for both upper and lower plates. Subduction rights the density inversion that is produced when asthenosphere is cooled to oceanic lithosphere: plate tectonics is driven by top-down cooling but is enabled by heat. Slabs sink more steeply than they dip and, if old and dense, are plated down on the 660 km discontinuity. Broadside-sinking slabs push all sublithosphere oceanic upper mantle inward, forcing rapid spreading in shrinking oceans. Down-plated slabs are overpassed by advancing arcs and plates, and thus transferred to enlarging oceans and backarc basins. Plate motions make sense in terms of this subduction drive in a global framework in which the ridge-bounded Antarctic plate is fixed: most subduction hinges roll back in that frame, plates move toward subduction zones, and ridges migrate to tap fresh asthenosphere. This self-organizing kinematic system is driven from the top. Slabs probably do not subduct into, nor do plumes rise to the upper mantle from, the sluggish deep mantle.