Continental displacements of thousands of kilometres point to flow patterns in the mantle of similar dimensions. As creep depends exponentially on temperature and as it is known that the temperature of the crust increases rapidly with depth, we can in this context suppose the mantle to have a sharp transition between a rigid crust and a fluid mantle at 50 to 100 km depth. Recently the coefficients of the tesseral harmonics of the geopotential from satellite observations have been determined. These departures from hydrostatic equilibrium seem to be caused by flow in the mantle, for from these coefficients can be computed the tractions exerted by the flow on the crust, assuming Navier-Stokes's equations, and the resulting pattern accords well with the world-wide tectonic features. Rising flow is associated with the East Pacific Ocean Rise, the Mid-Atlantic Ridge and the Mid-Indian Ocean Rise. Descending currents coincide with the Andes, the Alps and the Japan trench. The strong fifth harmonic in the satellite gravity data suggests that a flow pattern of this degree is at present establishing itself. On the other hand, continental reconstructions prior to Wegenerian drift suggest the presence of a strong fourth harmonic in the flow pattern. The theory of marginal stability in thermal convection, discussed by S. Chandrasekhar, shows that convection in a spherical shell under a uniform radial gravitational field gives the critical ratio of the radii of the inner to outer surfaces ($\eta $) at which the fifth order is more likely to develop than the fourth as 0$\cdot $54. The closeness of this value of $\eta $ to the present one of 0$\cdot $55 provides a clue to the most puzzling feature of continental drift; that it should have occurred in the last 5% of the Earth's history. H. C. Urey suggested that, on the accretion theory of the Earth's origin, the separation of iron towards the centre might have gradually occurred through the Earth's life. A growth of the core from 0$\cdot $54 to 0$\cdot $55 in the last 100-200 My does not conflict with the known rate of change in the length of the day. This explanation of the flow in terms of convection and changes in the degree of the flow pattern as a result of growth of the core implies earlier epochs of continental displacements, which are identified with peaks in the histograms of radiometric ages in the various continents. The law of growth of the core, so obtained, explains why so few rocks of greater age than 3000 My are found.