A limited comparison over the Northern Hemisphere oceans has been made between sea surface temperatures obtained from “Marine lkcks,” air temperatures over the ocean obtained from the same decks, and the historical file of hydrographic data. The intercomparison of these data suggest the following conclusions. 1) The SST observations have been contaminated by a systematic conversion from bucket to injection measurements. The bias so introduced may constitute as much as 30 to 50% of the observed change in sea surface temperature since the turn of the century. 2) The same bias effects are apparent in data sets that are alleged to contain bucket measurements of sea surface temperature only. 3) The behavior of the temperature field over the ocean appears to have significant and substantial differences from the behavior of estimated temperature changes over the Northern Hemisphere land masses. It seems clear that a reliable estimate of hemispheric or global temperature cannot be made without including adequate coverage of the ocean regions. 4) Many of the data suggest a shift in mean state of the sea surface temperature field of certain regions. It appears that this change is partially real and partially due to the merging of rather different types of data. 5) All estimates of pentad average temperature since 1900 are attended by relatively large standard deviations. This fact makes definitive discussion of pentad-to-pentad or decade-to-decade changes in hemispheric and global temperature difficult, if not impossible. Different methods of calculating the uncertainty in pentad averages could effect this conclusion.