Within practical limits of definition, the world's two most prominent volcanic hot spots, Hawaii and Iceland, are exactly 90 degrees apart. With them as control points and on the assumption that the Earth has dynamic symmetry at great depths, a global convection framework is devised, comprising three mutually perpendicular great circles intersecting at six convection centers. The Hawaiian antipode is near the Okavango delta in southern Africa, not far from the Bushveld Complex; the Iceland antipode is off the edge of Antarctica between the Balleny Islands and the Pacific-Antarctic ridge. The other two centers are at the south end of Peru and near the coast of Vietnam. Indications are that the centers at Hawaii, Iceland, Okavango, and Balleny overlie axes of deep upwelling in the Earth's convection system, while those at Peru and Vietnam are essentially regions of downwelling. Evidence comes from the lithospheric plate motions and broad tectonic relationships, from the tomography (thermal structure) of the deep mantle, from the dynamic topography of the core-mantle boundary, and from the geomagnetic field. The correlations with the magnetic field and core-mantle boundary suggest that the framework reflects convection in the liquid part of the core as well as the mantle. Present-day tectonic relationships around each of the framework centers are described with particular emphasis on justifying the tectonically relatively inactive Okavango site. Various parallels of history between the antipode centers of upwelling support the assumption that there is convective symmetry at depth, and they imply that this symmetry has been significant to Earth developments since at least the early Cenozoic. A prime example is that Australia separated from Antarctica at the Balleny center at almost the same time that Greenland and Europe separated around Iceland. Convection structures for the core and mantle are then devised. The core structure combines the convection framework with a fluid dynamic model based on the Coriolis force and with published patterns of core liquid motion derived from secular variations of the magnetic field. The structure would be expected to yield a magnetic field with at least some nondipolar characteristics, and it appears to have the potential to produce, not only frequent polarity reversals but also the long periods of stable polarity called superchrons. The mantle structure is based on a conceptual bilateral convergence model for the lower mantle and on stratified convection mechanisms for the upper mantle involving varied development of mechanical and thermal coupling of interfaces at the 400- and 670-km seismic discontinuities. The convergence model broadly accounts for the distribution of most of the world's currently active subduction zones, whereas the coupling mechanisms cover the main discrepancies and yield relationships whereby the deep mantle upwelling beneath Hawaii, Iceland, Balleny, and Okavango can drive the lithospheric spreading at the mid-ocean ridges. Findings of particular interest are that the upwelling beneath Hawaii could have powered the motion of the Pacific plate at the same time that it has been a hot spot magma source for the Emperor and Hawaiian seamount chains, and that continents can be held captive over an axis of upwelling when the coupling produces a system of mid-ocean ridges that encircles them both, as in the present-day circumstances of Africa over the Okavango axis and Antarctica next to the Balleny axis. The inferred distribution of thermal coupling correlates with the general global distribution of volcanic hot spots, and tomographic evidence is identified to support the postulate of mechanical coupling at 400 km. Certain global mantle geochemical features may also be accommodated . In the last part of the paper, the frequency variations of the geomagnetic polarity reversals since the Permian are shown to correlate closely with continental basalt floods and related rifting events, and on this basis, attempts are made to locate the convection framework on paleocontinental reconstructions of Pangea. Two perspectives are considered. By the one, the Okavango center is held fixed beneath southern Africa; by the other, the convection framework rotates rather freely within constraints relating to the events of continental rifting and volcanism. In both cases, the relationships suggest that the upwelling beneath the Okavango and Iceland centers largely controlled the formation of the giant Mesozoic and Cenozoic floods of continental basalt and their associated layered intrusions; these centers appear also to have played major roles in the origins of the African and Siberian fields of kimberlite, carbonatite, and other alkaline rocks, presumably (in part at least) by delivering large amounts of CO2 from the deep interior of the Earth to the continental lithosphere. The reconstructions for the moving framework additionally suggest that rifting effects along the paths of the centers of upwelling carved Pangea into the continental blocks that were dispersed during the Mesozoic and Cenozoic. The interaction or coupling of the framework with these blocks appears at times to have controlled its movements, and other magmatic hot spots may have been initiated (or rejuvenated) in the process.