The Early Cretaceous Ontong Java Plateau (OJP) represents by far the largest igneous event on Earth in the last 200 Ma and yet, despite its size, the OJP's basaltic crust appears to be remarkably homogeneous in composition. The most abundant rock type is a uniform low-K tholeiite, represented by the Kwaimbaita Formation on Malaita and found at all but one of the Deep Sea Drilling Project (DSDP) and Ocean Drilling Program (ODP) drill sites on the plateau and in the adjacent basins. This is capped by a thin and geographically restricted veneer of a slightly more incompatible-element-rich tholeiite (the Singgalo Formation on Malaita and the upper flow unit at ODP Site 807), distinguished from Kwaimbaita-type basalt by small but significant differences in Sr-, Nd- and Pb-isotope ratios. A third magma type is represented by high-Mg (Kroenke-type) basalt found in thick (>100 m) successions of lava flows at two drill sites (ODP Sites 1185 and 1187) 146 km apart on the eastern flank of the plateau. The high-Mg basalt is isotopically indistinguishable from Kwaimbaita-type basalt and may therefore represent the parental magma for the bulk of the OJP. Low-pressure fractional crystallization of olivine followed by olivine+augite+plagioclase can explain the compositional range from high-Mg Kroenke-type to Kwaimbaita-type basalt. The Singgalo-type basalt probably represents slightly smaller-degree, late-stage melting of an isotopically distinct component in the mantle source. Primary magma compositions, calculated by incremental addition of equilibrium olivine to aphyric Kroenke-type basalt glass, contain between 15.6% (in equilibrium with Fo (sub 90) ) and 20.4% (Fo (sub 92) ) MgO. Incompatible-element abundances in the primary OJP magma can be modelled by around 30% melting of a peridotitic primitive-mantle source from which about 1% by mass of average continental crust had previously been extracted. This large degree of melting implies decompression of very hot (potential temperature >1500 degrees C) mantle beneath very thin lithosphere. The initiation of an exceptionally large and hot plume head close to a mid-ocean ridge provides the best explanation for the size, homogeneity and composition of the OJP, but is difficult to reconcile with the submarine eruption of virtually all of the basalt so far sampled.