In the absence of melting, the small amounts of H₂O (100-300 ppm) in the MORB mantle would be totally dissolved in nominally anhydrous minerals and no hydrous phases would occur. These amounts of H₂O would lower the solidus only an insignificant amount. In contrast, the solidus temperature is a very weak function of the mount of CO₂, and even the smallest amount would reduce the solidus at P > 1.9 GPa by > 300°C. The consistent occurrence of vesicles in MORBs and the strong dominance of CO₂ in these vesicles indicates that at least some CO₂ exists in all or almost all of the mantle beneath ridges. The seismic low-velocity zone (LVZ) closely follows the ocean ridge system, and with rare exceptions (Iceland, Afar, and possibly a few other localities), it does not extend to depths greater than ~ 250 km. The minimum potential temperature (Tp) required for the generation of basalts is ~ 1240°C, a Tp that could not avoid producing carbonatitic melts at very low melt fractions in the depth range of ~ 65-300 km. This confirms earlier conclusions that partial melting occurs in the LVZ. For a Tp of 1240-1260°C, both the depth range of melting and depth of maximum melting (~ 70 km) predicted by the phase relations closely match the observed S velocity variations that define the LVZ beneath oceanic ridges. This supports the presence of CO₂ as the cause of melting and a low Tp range for almost all ridges. Here the carbonated lherzolite solidus is extrapolated to higher pressures in a way that minimizes the solidus-adiabat intersection and thereby maximizes the viability of high Tp values. Even with this conservative extrapolation, melting models that involve strongly variable potential temperatures ranging up to ~ 1450-1500°C (McKenzie and Bickle, 1988; Langmuir et al., 1992) imply melting at depths extending into the mantle transition zone for a carbonated lherzolite. For almost all ridge segments, shear wave tomography does not support melting at these extreme depths and is consistent only with low potential temperatures. If the Azores, Galapagos, and Tristan "hot spots" are hot, their broad topographic swells extending over > 2000 km suggest the existence of deep and broad low shear velocity anomalies that should be easily visible by global seismic tomography. The absence of such anomalies supports low potential temperatures also for these near-ridge "hot spots".