a b s t r a c tFor more than half a century, thought about granite genesis and crustal evolution has been guided by the concept of partial melting in the lower crust. In this model, granitic magmas produced at depth are lost to shallow levels, leaving behind a more mafic, volatile poor residue that is depleted in incompatible components (H 2 O, alkalis, and heat-producing elements). Although granite extraction must be the dominant process by which crust is modified over time, the preferred model of granite genesis triggered by metamorphic dehydration reactions (dehydration melting) does not adequately explain important aspects of granite formation. The temperatures required for voluminous granite production by dehydration melting need heat and mass input to the crust from mantle-derived mafic magmas. In addition, prediction of the H 2 O contents of granitic liquids by extrapolation from low-pressure experiments to deep-crustal pressures (P) and temperatures (T) implies that the H 2 O resident in hydrous minerals is insufficient to account for large granite volumes, such as anorogenic granite batholiths in continental interiors. To test this, we conducted new experiments on the H 2 O contents of simple granitic liquids at 10 kbar and 800-950• C. We confirm previous extrapolations from lower P and T indicating that a minimum of 3-4 wt% H 2 O is present at the studied P and T in a granitic liquid in equilibrium with quartz and feldspars. For large-scale melting, this is much more than could have been supplied by the H 2 O resident in biotite and amphibole by dehydration melting at these conditions, unless lower-crustal temperatures were higher than generally inferred. Another problem with the dehydration-melting model is that the crystal chemistry of the large-ion lithophile elements (LILE) does not favor their partitioning into granitic liquids; rather, U, Th, Rb and the rare earth elements (REE) would more likely be concentrated in the postulated mafic residues. Finally, observations of migmatite complexes reveal many features that can not be satisfied by a simple dehydration-melting model.We suggest that the volatile components CO 2 and Cl are important agents in deep-crustal metamorphism and anatexis. They induce crystallization and outgassing of basalt magmas at lower-crustal levels, where the combination of latent heat and liberated H 2 O may contribute to granite production, leading to larger melt fractions than for simple dehydration-melting models. Since the Cl and CO 2 are very insoluble in granite liquids, granite generation leads naturally to production or separation of a coexisting metamorphic fluid with low H 2 O activity. Such a fluid could coexist with granulite-facies assemblages and yet be capable of dehydration, alkali exchange and LILE extraction to explain many chemical processes of deep-crustal metamorphism not readily explainable by dehydration melting.