A scheme of deep mantle convection is proposed in which narrow plumes of deep material rise and then spread out radially in the asthenosphere. These vertical plumes spreading outward in the asthenosphere produce stresses on the bottoms of the lithospheric plates, causing them to move and thus providing the driving mechanism for continental drift. One such plume is beneath Iceland, and the outpouring of unusual lava at this spot produced the submarine ridge between Greenland and Great Britain as the Atlantic opened up. It is concluded that all the aseismic ridges, for example, the Walvis Ridge, the Ninetyeast Ridge, the Tuamotu Archipelago, and so on, were produced in this manner, and thus their strikes show the direction the plates were moving as they were formed. Another plume is beneath Hawaii (perhaps of lesser strength, as it has not torn the Pacific plate apart), and the Hawaiian Islands and Emperor Seamount Chain were formed as the Pacific plate passed over this “hot spot.” Three studies are presented to support the above conclusion. (1) The Hawaiian-Emperor, Tuamotu-Line, and Austral-Gilbert-Marshall island chains show a remarkable parallelism and all three can be generated by the same motion of the Pacific plate over three fixed hot spots. The accuracy of the fit shows that the hot spots have remained practically fixed relative to one another in this 100 m.y. period, thus implying a deep source below the asthenosphere. (2) The above motion of the Pacific plate agrees with the paleo-reconstruction based on magnetic studies of Pacific seamounts. The paleomotion of the African plate was deduced from the Walvis Ridge and trends from Bouvet, Reunion, and Ascension Islands. This motion did not agree well with the paleomagnetic studies of the orientation of Africa since the Cretaceous; however, better agreement with the paleomagnetic studies of Africa and of seamounts in the Pacific can be made if some polar wandering is permitted in addition to the motion of the plates. (3) A system of absolute plate motions was found which agrees with the present day relative plate motions (deduced from fault strikes and spreading rates) and with the present trends of island chains-aseismic ridges away from hot-spots. This shows that the hot spots form a fixed reference frame and that, within allowable errors, the hot spots do not move about in this frame.