Examination of glass and crystal chemistry in the Rotoiti Pyroclastics (>100 km 3 of magma) demonstrates that compositional diversity was produced by mingling of the main rhyolite magma body with small volumes of other magmas that had been crystallizing in separate stagnant magma chambers. Most (>90%) of the Rotoiti deposits were derived from a low-K 2O, cummingtonite-bearing, rhyolitic magma (T1) discharged throughout the eruption sequence. T1 magma is homogeneous in composition (melt SiO 2=77.80±0.28 wt.%), temperature (766±13 °C) and oxygen fugacity (NNO+0.92±0.09). Most T1 phenocrysts formed in a shallow (∼200 MPa), near water-saturated ( a water=0.8) storage chamber shortly before eruption. Basaltic scoria erupted immediately before the rhyolites, and glass-bearing microdiorite inclusions within the rhyolite deposits, suggest that basalt emplaced on the floor of the chamber drove vigorous convection to produce the well-mixed T1 magma. Lithic lag breccias contain melt-bearing biotite granitoid inclusions that are compositionally distinct from T1 magma. The breccias which overlie the voluminous T1 pyroclastic flow deposits resulted from collapse of the syn-Rotoiti caldera. Post-collapse Rotoiti pumices contain T1 magma mingled with another magma (T2) that is characterized by high-K glass and biotite, and was cooler and less oxidised (712±16 °C; NNO−0.16±0.16). The mingled clasts contain bimodal disequilibrium populations of all crystal phases. The granitoid inclusions and the T2 magma are interpreted as derived from high-K magma bodies of varying ages and states of crystallization, which were adjacent to but not part of the large T1 magma body. We demonstrate that these high-K magmas contaminated the erupting T1 magma on a single pumice clast scale. This contamination could explain the reported wide range of zircon U–Th ages in Rotoiti pumices, rather than slow crystallization of a single large magma body.