Abstract:Typically melting occurs during decompression in ultra-high-pressure terranes, along the evolution from peak pressure to peak temperature in medium-temperature eclogitehigh-pressure granulite terranes and by simple prograde heating in granulite facies and ultrahigh-temperature (UHT) metamorphic terranes. The source of heat must be due to one or more processes among thickening and radiogenic heating, viscous dissipation, and heat from asthenospheric mantle. Melt-bearing rocks become porous at a few vol% melt initiating an advective flow regime. As the melt volume approaches and exceeds the melt connectivity transition (~ 7 vol% melt), melt may be lost from the system in the first of several melt-loss events. In migmatites and residual granulites, a variety of microstructures indicates the former presence of melt at the grain scale and leucosome networks at outcrop scale record melt extraction pathways. This evidence supports a model of focused melt flow by dilatant shear failure at low melt volume as the crust weakens with increasing melt production. Melt ascent is initiated as ductile fractures but continues in dykes that propagate as brittle fractures. Crustal rocks undergo melting via a sequence of reactions beginning with minimal melt production at the wet solidus (generally < 1 vol% melt, unless there is influx of H 2 O-rich fluid). The major phase of melt production is related to hydrate-breakdown melting (perhaps > 50 vol% melt, depending on the fertility of the protolith composition and the intensive variables). At temperatures above the stability of the hydrate, assuming significant melt loss by this point, low-volume melt production continues by consumption of feldspar(s) and quartz at UHT conditions (generally < 10 vol% melt at peak UHTM conditions). Significant melt loss is a contributory factor to achieving UHTs because dehydration of the system limits the progress of heat-consuming melting reactions among the residual phases in the source.