Don Anderson emphasizes that plate tectonics is self-organizing and is driven by subduction, which rights the density inversion generated as oceanic lithosphere forms by cooling of asthenosphere from the top. The following synthesis owes much to many discussions with him. Hinge rollback is the key to kinematics, and, like the rest of actual plate behavior, is incompatible with bottom-up convection drive. Subduction hinges (which are under, not in front of, thin leading parts of arcs and overriding plates) roll back into subducting plates. The Pacific shrinks because bounding hinges roll back into it. Colliding arcs, increasing arc curvatures, back-arc spreading, and advance of small arcs into large plates also require rollback. Forearcs of overriding plates commonly bear basins which preclude shortening of thin plate fronts throughout periods recorded by basin strata (100 Ma for Cretaceous and Paleogene California). This requires subequal rates of advance and rollback, and control of both by subduction. Convergence rate is equal to rates of rollback and advance in many systems but is greater in others. Plate-related circulation probably is closed above 650 km. Despite the popularity of concepts of plumes from, and subduction into, lower mantle, there is no convincing evidence for, and much evidence against, penetration of the 650 in either direction. That barrier not only has a crossing-inhibiting negative Clapeyron slope but also is a compositional boundary between fractionated (not “primitive”), sluggish lower mantle and fertile, mobile upper mantle. Slabs sink more steeply than they dip. Slabs older than about 60 Ma when their subduction began sink to, and lie down on and depress, the 650-km discontinuity, and are overpassed, whereas younger slabs become neutrally buoyant in mid-upper mantle, into which they are mixed as they too are overpassed. Broadside-sinking old slabs push all upper mantle, from base of oceanic lithosphere down to the 650, back under shrinking oceans, forcing rapid Pacific spreading. Slabs suck forward overriding arcs and continental lithosphere, plus most subjacent mantle above the transition zone. Changes in sizes of oceans result primarily from transfer of oceanic lithosphere, so backarcs and expanding oceans spread only slowly. Lithosphere parked in, or displaced from, the transition zone, or mixed into mid-upper mantle, is ultimately recycled, and regional variations in age of that submerged lithosphere may account for some regional contrasts in MORB. Plate motions make no kinematic sense in either the “hotspot” reference frame (HS; the notion of fixed plumes is easily disproved) or the no-net-rotation frame (NNR) In both, for example, many hinges roll forward, impossible with gravity drive. Subduction-drive predictions are fulfilled, and paleomagnetic data are satisfied (as they are not in HS and NNR), in the alternative framework of propulsionless Antarctica fixed relative to sluggish lower mantle. Passive ridges migrate away from Antarctica on all sides, and migration of these and other ridges permits tapping fresh asthenosphere. (HS and NNR tend to fix ridges). Ridge migration and spreading rates accord with subduction drive. All trenches roll back when allowance is made for back-arc spreading and intracontinental deformation. Africa rotates slowly toward subduction systems in the NE, instead of moving rapidly E as in HS and NNR. Stable NW Eurasia is nearly stationary, instead of also moving rapidly, and S and E Eurasian deformation relates to subduction and rollback. The Americas move Pacificward at almost the full spreading rates of passive ridges behind them. Lithosphere has a slow net westward drift.

Reference: W.B. Hamilton, An alternative Earth, GSA Today, in press.