Two-stage diapiric models for lunar ferroan anorthosites and terrestrial massif anorthosites are examined. The lunar model is developed to explain early lunar differentiation in the absence of a magma ocean. If correct, the terrestrial model serves as an analog for the development of lunar anorthositic diapirs. There is field and textural evidence of transport of mostly crystalline anorthositic material within the terrestrial complexes. This evidence, combined with the absence of anorthositic lavas and phase equilibrium constraints inhibiting the production of hyperaluminous magmas, is consistent with the detachment of plagioclase-rich crystalline mushes from large, uppermost mantle plutons and multiple diapiric intrusion of these mushes into the upper sialic crust with attendant anatexis. Rudimentary dynamical calculations suggest that a simple, single-layer source of the diapirs is improbable: either there were several parental magma chambers or there was a single large chamber that was repeatedly replenished. The lunar model is a development of Wetherill's (1975) suggestion that, following accretion, the outer portion of the moon consisted of a stack of overlapping layered intrusions and that reheating of these intrusions mobilized their anorthositic layers, which intruded upwards to produce the anorthositic lunar crust. Dynamical calculations show that gravitational instabilities in anorthositic layers (≤1 km thick) could develop into diapirs on a reasonable time scale (50--100 m.y.) only if the outer portion of the moon was partially remelted. We suggest that partial melting of the interior due to the decay of long-lived radionuclides with the subsequent onset of global convection heated the stack of intrusions from below, thus causing melting and allowing the anorthositic diapirs to grow and ascend fairly rapidly. As the convecting zone thickened, a mass expulsion of anorthositic material out of the lunar interior may have occurred.