Geochemical data have been interpreted as requiring that a significant fraction of the melting in MORE source regions takes place in the garnet peridotite field, an inference that places the onset of melting at greater than or equal to 80 km. However, if melting begins at such great depths, most models for melting of the suboceanic mantle predict substantially more melting than that required to produce the 7 +/- 1 km thickness of crust at normal ridges. One possible resolution of this conflict is that MORBs are produced by melting of mixed garnet pyroxenite/spinel peridotite sources and that some or all of the ''garnet signature'' in MORE is contributed by partial melting of garnet pyroxenite layers or veins, rather than from partial melting of garnet peridotite. Pyroxenite layers or veins in peridotite will contribute disproportionately to melt production relative to their abundance, because partial melts of pyroxenite will be extracted from a larger part of the source region than peridotite partial melts (because the solidus of pyroxenite is at lower temperature than that of peridotite and is encountered along an adiabat 15-25 km deeper than the solidus of peridotite), and because melt productivity from pyroxenite during upwelling is expected to be greater than that from peridotite (pyroxenite melt productivity will be particularly high in the region before peridotite begins melting, owing to heating from the enclosing peridotite). For reasonable estimates of pyroxenite and peridotite melt productivities, 15-20% of the melt derived from a source region composed of 5% pyroxenite and 95% peridotite will come from the pyroxenite. Most significantly, garnet persists on the solidus of pyroxenite to much lower pressures than those at which it is present on the solidus of peridotite, so if pyroxenite is present in MORE source regions, it will probably contribute a garnet signature to MORE even if melting only occurs at pressures at which the peridotite is in the spinel stability field. Partial melting of a mixed spinel peridotite/garnet pyroxenite mantle containing a few to several percent pyroxenite can explain quantitatively many of the geochemical features of MORE that have been attributed to the onset of melting in the stability field of garnet Iherzolite, provided that the pyroxenite compositions are similar to the average composition of mantle-derived pyroxene-rich rocks worldwide or to reasonable estimates of the composition of subducted oceanic crust. Sm/Yb ratios of average MORE from regions of typical crustal thickness are difficult to reconcile with derivation by melting of spinel peridotite only, but can be explained if MORE sources contain similar to 5% garnet pyroxenite. Relative to melting of spinel peridotite alone, participation of model pyroxenite in melting lowers aggregate melt Lu/Hf without changing Sm/Nd ratios appreciably. Lu/Hf-Sm/Nd systematics of most MORE can be accounted for by melting of a spinel peridotite/garnet pyroxenite mantle provided that the source region contains 3-6% pyroxenite with greater than or equal to 20% modal garnet. However, Lu/Hf-Sm/Nd systematics of some MORE appear to require more complex melting regimes and/or significant isotopic heterogeneity in the source. Another feature of the MORE garnet signature, (Th-230)/(U-238) > 1, can also be produced under these conditions, although the magnitude of (Th-230)/(U-238) enrichment will depend on the rate of melt production when the pyroxenite first encounters the solidus, which is not well-constrained. Preservation of high (Th-230)/(U-238) in aggregated melts of mixed spinel peridotite/garnet pyroxenite MORE sources is most likely if the pyroxenites have U concentrations similar to that expected in subducted oceanic crust or to pyroxenite from alpine massifs and xenoliths. The abundances of pyroxenite in a mixed source that are required to explain MORE Sm/Yb, Lu/Hf, and (Th-230)/(U-238) are all similar. If pyroxenite is an important source of garnet signatures in MORBI then geochemical indicators of pyroxenite in MORE source regions, such as increased trace element and isotopic variability or more radiogenic Pb or Os, should correlate with the strength of the garnet signature. Garnet signatures originating from melts of the garnet pyroxenite components of mixed spinel peridotite/garnet pyroxenite sources would also be expected to be stronger in regions of thin crust.
