In order to understand the role of mica-rich rocks as a source of granite magmas, a series of melting experiments was performed on two different starting materials. The first composition is a model biotite gneiss consisting of 30 sw % biotite, 30 wt % plagioclase, and 40 wt % quartz. The second composition is a model two-mica pelites consisting of 15 wt % biotite, 15 wt % muscovite, 30 wt % plagioclase, and 40 wt % quartz. Experiments were performed under vapor-absent conditions at 1.0 GPa and between 750¿ and 950 ¿C. With only biotite in the starting material the volume of melt is always less than 15 vol % below 900 ¿C and reaches 25 vol % at 950 ¿C. In experiments that involve both biotite and muscovite in the starting material, the melt proportion increases up to 28 vol % at 825 ¿C and reaches 60 vol % at 950 ¿C. For the biotite-plagioclase-quartz (BPQ) assemblage, the solidus is located at 800 ¿C at 1.0 GPa. The melting reaction produces a metaluminous granitic liquid and leaves a residuum consisting of garnet+biotite+orthopyroxene+plagioclase+quartz. In addition, the experiments show that at 1.0 GPa biotite can be stable above 950 ¿C. With both micas in the starting material (BPQM), the solidus at 1.0 GPa is located at 750 ¿C. The melting reactions produce a peraluminous granitic liquid and leave a residuum of garnet+sillimanite+biotite+quartz+plagioclase+Kfeldspar in experiments below 900 ¿C. At 950 ¿C the residuum consists of garnet+orthopyroxene+biotite+plagioclase. The melt fraction is determined by the proportions of the hydrous phases and of the amount of feldspar relative to quartz. Mineral modes of the source rocks, particularly the amount of quartz, are at least as important as the amount of available H2O in controlling the melt fraction generated during crustal anatexis. ¿ American Geophysical Union 1995