Experiments at 20-97 GPa and 2000 K in the system CaO-MgO-TiO. 2-SiO. 2 constrain phase relations involving Mg-rich and Ca-rich perovskite solid solutions at conditions relevant to the Earth's deep Transition Zone and lower mantle. Bulk compositions were investigated with molar Ti/(Ti + Si) up to 0·5 within the quasi-ternary 'perovskite plane', which is defined by a reciprocal solution among the components MgSiO 3, MgTiO 3, CaSiO 3, and CaTiO 3. Multi-anvil experiments at 20 GPa and 2000 K on bulk compositions within the plane produce akimotoite coexisting with Ca-perovskites that lie close to the CaSiO 3-CaTiO 3 join. Higher-pressure experiments using a laser-heated diamond anvil cell constrain the position of a two-perovskite field that extends into the perovskite plane from the solvus along the MgSiO 3-CaSiO 3 binary join, where limited mutual solubility exists between MgSiO 3 and CaSiO 3 perovskites. On the join MgSiO 3-MgTiO 3, MgTiO 3 solubility in MgSiO 3 perovskite increases with pressure, with MgSi 0·8Ti 0·2O 3 perovskite stable at ∼50 GPa. Limited reciprocal solution at ∼25 GPa results in an expansive two-perovskite field that occupies much of the Si-rich portion of the perovskite plane. Solution of Ti into Mg-rich and Ca-rich perovskites enhances the solubility of reciprocal Ca and Mg components, respectively. Increase in pressure promotes reciprocal solution, and the two-phase field collapses rapidly with pressure toward the MgSiO 3-CaSiO 3 join. We find that a single-phase, orthorhombic perovskite with near equimolar Ca and Mg is stable in a composition with Ti/(Ti + Si) of only 0·05 at 97 GPa, requiring that by this pressure the two-phase field occupies a small area close to the MgSiO 3-CaSiO. 3 join. On the basis of experiments at∼1500 K, temperature has only a mild effect on the position of the Ca-rich limb of the solvus. Ca(Ti,Si)O 3 mineral inclusions in deep sublithospheric diamonds could not have formed in equilibrium with Mg-perovskite owing to their virtual lack of MgSiO 3 component at pressures of Mg-perovskite stability, but may have equilibrated with Transition Zone MgSiO 3-rich phases at lower pressures; this observation can be extended generally to near-endmember CaSiO 3 inclusions. On an iron-free basis, the average bulk compositions of clinopyroxene-ilmenite and orthopyroxene-ilmenite megacrysts from kimberlites plot in single-perovskite fields at pressures greater than about 45 and 65 GPa, respectively, when projected onto the perovskite plane. We predict that the effect of iron will not be large, and estimate that single-phase perovskites may form at somewhat lower pressures than in the iron-free system. Thus, the origin of pyroxene-ilmenite megacrysts from single-phase perovskite solutions in the lower mantle is plausible on the basis of phase relations, although a lower pressure magmatic origin appears more likely. Deeply subducted Ti-rich lithologies such as ocean-island basalt will crystallize a single perovskite rather than a two-perovskite assemblage beginning at pressures of ∼80 GPa. Normal mid-ocean ridge basalt and primitive mantle peridotite are expected to remain within a two-phase perovskite field until Mg-perovskite transforms to post-perovskite.