By treating the lithosphere as a diffusive boundary layer to mantle convection, the convective speed or mantle creep rate, ϵ ̇ , can be related to the mantle-derived heat flux, Q ̇ . If cell size is independent of Q ̇ 2 then ϵ ̇ ∝ Q ̇ . (If cell size varies with Q ̇ , then a different power law prevails, but the essential conclusions are unaffected.) Then the factthat for constant thermodynamic efficiency of mantle convection, the mechanical power dissipation is proportionalto Q ̇ , gives convective stress σ ∝ Q ̇ −1 , i.e. the stress increases as the convection slows. This means an increasing viscosityor stiffness of the mantle which can be identified with a cooling rate in terms of a temperature-dependent creep law. If we suppose that the mantle was at or close to its melting point within 1 or 2 × 10 8 years of accretionof the Earth, the whole scale of cooling is fixed. The present rate of cooling is estimated to be about 4.6 × 10 −8 deg y −1 for the average mantle temperature, assumed to be 2500 K, but this very slow cooling rate represents a loss ofresidual mantle heat of 7 × 10 12 W, about 30% of the total mantle-derived heat flux. This conclusion requires theEarth to be deficient in radioactive heat, relative to carbonaceous chondrites. A consideration of mantle outgassing and atmospheric argon leads to the conclusion that the deficiency is due to depletion of potassium, and that the K/U ratio of the mantle is only about 2500, much less than either the crustal or carbonaceous chondritic values. Thetotal terrestrial potassium is estimated to be about 6 × 10 20 kg. Acceptance of the cooling of the Earth removes the necessity for potassium in the core.