We present a new method for calculating the major element compositions of primary melts parental to mid-ocean ridge basalt (MORB). This model is based on the experimental data of Jaques and Green (1980), Falloon et al. (1988), and Falloon and Green (1987, 1988) which are ideal for this purpose. Our method is empirical and employs solid-liquid partition coefficients (Di) from the experiments. We empirically determine Di=f(P,F) and use this to calculate melt compositions produced by decompression-induced melting along an adiabat (column melting). Results indicate that most MORBs can be generated by 10--20% partial melting at initial pressures (P0) of 12-21 kbar. Our primary MORB melts have MgO=10-12 wt %. We fractionate these at low pressure to an MgO content of 8.0 wt % in order to interpret natural MORB liquids. This model allows us to calculate Po, Pf, To, Tf, and F for natural MORB melts. We apply the model to interpret MORB compositions and mantle upwelling patterns beneath a fast ridge (East Pacific Rise (EPR) 8¿N to 14¿N), a slow ridge (mid-Atlantic Ridge (MAR) at 26¿S), and seamounts near the EPR (Lamont seamount chain). We find mantle temperature differences of up to 50¿--60 ¿C over distances of 30--50 km both across axis and along axis at the EPR. We propose that these are due to upward mantle flow in a weakly conductive (versus adiabatic) temperature gradient. We suggest that the EPR is fed by a wide (~100 km) zone of upwelling due to plate separation but has a central core of faster buoyant flow. An along-axis thermal dome between the Siqueiros transform and the 11¿45' Overlapping Spreading Center (OSC) may represent such an upwelling; however, in general there is a poor correlation between mantle temperature, topography, and the segmentation pattern at the EPR.

For the Lamont seamounts we find regular across-axis changes in Po and F suggesting that the melt zone pinches out off axis. This observation supports the idea that the EPR is fed by a broad upwelling which diminishes in vigor off axis. In contrast with the EPR axis, mantle temperature correlates well with topography at the MAR, and there is less melting under offsets. The data are consistent with weaker upwelling under offsets and a adiabatic temperature gradient in the subaxial mantle away from offsets. The MAR at 26¿S exhibits the so-called local trend of Klein and Langmuir (1989). Our model indicates that the local trend cannot be due solely to intracolumn melting processes. The local trend seems to be genetically associated with slow-spreading ridges, and we suggest it is due to melting of multiple individual domains that differ in initial and final melting pressure within segments fed by buoyant focused mantle flow. ¿ American Geophysical Union 1991