Planetary crusts may be accreted, produced in primordial differentiation, or built up piecemeal by serial magmatism. The existence of old, polygenetic, laterally heterogeneous, partial melt rocks in the lunar highlands suggests that the moon produced its early crust by serial magmatism. This view can be reconciled with lunar Eu anomalies, previously thought to support the magma ocean model of crust formation, if complications in the fractionation of mare basalts are reconized. Phase equilibrium and magmatic density information for mare basalts suggest a model in which plagioclase fractionation can occur even though plagioclase is not a near-liquidus phase. The crytic fractionation of clinopryoxene in MORB provides a precedent for this model. The necessity for a lunar magma ocean is questioned, but a role for a terrestrial magma ocean of sorts at depth is suggested.