Useful constraints on the depth-temperature range and degree of melting of lower crustal rocks have been deduced by combining the results from experimental petrology with plausible PTt paths calculated for continental collision zones. H2O is required to generate melts at the temperatures of crustal orogenesis. Our modeling has shown that residual free water from subsolidus dehydration reactions produces less than 0.5% granitic melt at the H2O-saturated solidus. Breakdown of hydrous minerals results in fluid-absent (dehydration)-melting. For average geotherms such melting of muscovite and biotite in metapelite will generate at most 25% granitic melt. Because of much higher solidus temperatures, fluid-absent melting of hornblende in metabasalt (amphibolite) will not occur in collision zones without augmented heat input from the mantle. Fluid-absent melting of epidote in metabasalt, and low Ca amphibole and biotite in metavolcanics, can produce up to 10% melt in continental collision zones. Minor melt production from dehydration-melting of chlorite, staurolite, chloritoid, or talc will occur at pressures greater than ~0.6 GPa. The calculated paths envelope small regions of PT space, thus identifying which mineral reactions are important for crustal anatexis and at which PT conditions future experiments should be made. In a dynamically evolving crust during uplift, thermal buffering by melting reactions is a minor component of the heat budget. Extension of thickened continental crust will not promote dehydration-melting without the incursion of mantle heat. Delamination of thickened eclogitic lower crust, or decompression of the asthenosphere in response to lithosphere thinning are the most promising mechanisms to induce extensive lower crustal melting including amphibolite. ¿ American Geophysical Union 1995