The thermal structure of the oceanic upper mantle is examined on the basis of seismic attenuation, surface heat flow, and mid-ocean ridge basalt (MORB) composition. The comparison of seismic Q with laboratory Q determined at sufficiently high pressure and temperature allows us to constrain temperature in the Earth. Since the inferred absolute temperature relies on knowledge of the frequency dependence of Q (if any), we calibrate an unknown frequency effect on Q by comparing the derived temperatures with geotherms calculated from surface heat flow and MORB composition. This comparison yields a correction factor (found to be small) for the geotherm derived from laboratory (ultrasonic) Q. Examples of thermal structure are presented for the Iceland Plateau and the Pacific Ocean. Under the Iceland Plateau, we expect partial melt for the asthenosphere only in the vicinity of the spreading ridge. In asthenosphere older than ~5 m.y. of the Iceland Plateau and older than ~40 m.y. of the Pacific Ocean, temperatures are below the dry peridotite solidus, and no partial melt is expected. These results are consistent with the thermal structure derived from a comparison of seismic and laboratory velocities. It is shown that there is a good agreement between temperatures determined from seismic attenuation, heat flow,, and MORB chemistry. Since the laboratory measurements were made on a dry peridoties and we employ a dry peridotite solidus for determining the geotherm, the consistency in temperatures derived from three independent data strongly suggests that the upper mantle (remote from subduction zones) is essentially dry. It appears that the spreading process in mid-ocean ridge causes asthenosphere material to ascend from greater depths where it was at higher temperature. This give rise to pressure release melting. ¿ American Geophysical Union 1989