Density and elasticity data are consistent with, but do not require, a uniform upper‐ and lower‐mantle composition. Such data cannot at present resolve the small changes in physical properties that would be required to keep distinct mantle reservoirs dynamically separated. Alternatively, the issue of chemical stratification in the mantle can be addressed by analyzing the details of the phase equilibria of the appropriate minerals. In order to do this we construct an internally consistent pressure calibration scale which is based in part on the coesite‐stishovite transition pressurePtr(GPa) = 7.45(±0.3) + 2.1(±0.3) × 10−3T(°C) according to our reanalysis of the available data. We conclude that a discontinuous reaction occurs in the olivine component (α+γ = α+β) of the mantle at the conditions of the 400‐km seismological discontinuity; however, no discontinuous reactions corresponding to the 670‐km discontinuity have yet been identified. The only reaction observed in diamond cell experiments at approximately the pressures existing at 670‐km depth, the breakdown of γ‐spinel to form a silicate perovskite assemblage, appears not to satisfy the observed sharpness of this discontinuity. Thus it may be necessary to invoke either a univariant reaction that has not yet been observed experimentally or a chemical discontinuity at this depth. As the mantle is likely to be at temperatures higher than those of the experiments (estimated to be near 1000°C), additional univariant reactions involving the silicate ilmenite or garnet phases that are predicted to occur may be significant. One possible interpretation is that the mantle is of uniform composition and that such (hypothetical) reactions produce the 670‐km discontinuity. This would imply that the phase assemblages so far studied in high‐pressure‐temperature experiments are not those occurring in the mantle. Alternatively, the mantle