Rabaul caldera is unusual in that it was formed by two episodes of construction and collapse on an older basalt volcano. One collapse occurred around 3,500 yr B.P. and the latest around 1,400 yr B.P. Both were accompanied by the eruption of dacitic pumiceous ash flows. Following initial collapse, volcanism was confined to a large andesite volcano in the southern part of the caldera, but a renewal of basaltic volcanism occurred at a point on the eastern ring fracture. The later collapse of the southern cone left the caldera as a large ellipse, 14 by 9 km, breached on the southeast side, though more recent volcanism has built small cones on the caldera floor.
The rocks of the caldera cannot unequivocally be labeled as calc-alkalic because the contemporaneous basalts are high-alumina types and the suite shows mild iron enrichment. This mild iron enrichment is not found in the lavas of some parasitic volcanoes, suggesting that it is an artifact of the magma chamber beneath the caldera. Differences in chemistry and mineralogy persist between lava erupted on the western side of the caldera and that on the east; chemical variation is explained by crystal fractionation at low pressures. Lack of a clearly defined Benioff zone and location close to a transform fault suggest that the origin of the magma was not controlled by partial melting near a Benioff zone. Comparison with experimental data for high-alumina basalt also suggests a shallower origin. A progressive increase in TiO2, Al2O3, and alkalis along the New Britain–New Guinea arc toward Rabaul at its eastern end cannot be explained.