Calculation of the volume and composition of magma generated by lithospheric extension requires an accurate mitial geotherm, and knowledge of the variation and composition of the melt fraction as a function of pressure and temperature. The relevant geophysical observations are outlined, and geotherms then obtained from parameterized convective models. Experimental observations which constrain the solidus and liquidus at various pressures are described by simple empirical functions. The variation in melt fraction is then parameterized by requiring a variation from 0 on the solidus to 1 on the liquidus. The composition of the melts is principally controlled by the melt fraction, though those of FeO, MgO, and SiO2 in addition vary with pressure. Another simple parameterization allows the observed compositions of major elements in 91 experiments to be calculated with a mean error of 1.1%, and those of TiO2 and Na2O to 0.3%. These expressions are then used to calculate the expected compositions of magma produced by adiabatic upwelling. The mean major element composition of the most magnesium-rich MORB glasses resemble the mean composition calculated for a mantle potential temperature Tp of 1280°C. Adiabatic melting during upwelling of mantle of this temperature generates a melt thickness of 7 km. The observed variations of the MgO and TiO2 concentrations in a large collection of MORB glass compositions suggest that extensive low pressure fractional crystallization occurs, but that its effect on the concentrations of SiO2, Al2O3, and CaO is small. There is no evidence that normal oceanic crust is produced from magmas containing more than 11% MgO. The mantle potential temperature within hot rising jets is about 1480°C and can generate 27 km of magma containing 17% MgO. Extension of the continental lithosphere generates little melt unless β> 2 and Tp> 1380°C. The melts generated by larger values of β or of Tp are alkali basalts, and change to tholeiites as the amount of melting increases. Large quantities of melt can be generated, especially at continental margins, where estimates of β obtained from changes in crustal thickness will in general be too small.