In this paper we elaborate on suggestions in the recent literature that the mantle convects uniformly through its entire depth. The main novel feature introduced here, which leads to a satisfactory account of earth temperatures, it the assumption of a thermal boundary layer at the bottom of the mantle. Such a layer is produced by the contrast of thermal conductivities of core and mantle if, as is now generally believed, the effects of radiative heat transfer in the mantle are small. On assuming in agreement with much current geochemical thinking that the core is mainly a solution of FeS in Fe, it is possible to estimate crudely the temperature of the core-mantle boundary as 4000¿500¿K. The temperature curve in the mantle can be approximated by adding two boundary layer temperature drops that can be calculated to the adiabat in between: this also calculable, and the sum of these three agrees roughly with the numerical value just given. It has long been known that the Rayleigh number of the mantle is so large as to make convection likely. Lately, Golitsyn has introduced a scaling analysis that allows one to express the depth of the convective zone as a function of known parameters: this yields a depth in rough agreement with the depth of the entire mantle. Finally, we discuss the likelihood that mantle convection has been going on through the entire life of the earth, beginning with the early formation of the core; this has obvious geological implications.