The development of reliable paleoclimatic maps at a global scale requires data at the following three levels of analysis: (1) well-recorded observations of evenly positioned, well-dated geological evidence (Level I), (2) paleoclimatic estimates derived from this evidence by well-defined quantitative repeatable methods (Level II), and (3) maps synthesizing the estimates from several independent sources of geological evidence (Level III). Our paper describes much of the currently available paleoclimatic data from unglaciated terrestrial areas at ca. 18,000 yr B.P. and illustrates the quantity and quality of the data at both the Level I and the Level II stages of analysis. Although the scarcity of well-dated evidence for this time period precluded any major Level III syntheses of the information, comparisons were drawn where possible between the geological evidence and the climatic conditions simulated by general-circulation model experiments of Gates (1976a, b) and Manabe and Hahn (1977). Of the more than 320 sites with data from 18,000 yr B.P., only 65 are well-dated with bracketing dates within the interval of 23,000 to 13,000 yr B.P., whereas about 100 are undated or poorly dated. We concentrated our survey on palynological and paleobotanical evidence and also thoroughly reviewed the evidence for water levels in lakes at 18,000 yr B.P. In areas with few of these sources of evidence, data on former snowlines, periglacial features, and eolian deposits were included, but the survey of these data is far from complete. Maps of the assembled data reveal the consistency of the paleoclimatic estimates in “data-rich” areas and also show which areas required additional information. The maps show that conditions were colder than present at 18,000 yr B.P. for all sites with temperature estimates. Estimated temperature depressions varied from ca. 1° to 12°C or more, depending on the location of the sample, the type of geological evidence, and the method of temperature estimation. Interpreted hydrological conditions were more variable spatially than the temperature estimates. The southwestern U.S. was moister than present, whereas the southeast may have been drier. Europe and the northern Mediterranean across to Afghanistan were drier than present, but northwest Africa was wetter. Australia was mainly drier than present, but several sites there as well as in Africa show significant climatic changes between 21,000 year period and 16,000 yr B.P. This latter evidence suggests that considerable variability may have occurred during the several thousand-centered on 18,000 yr B.P. Accurate time control is therefore required for the geological data used to study the climate dynamics of 18,000 yr ago. Large portions of South America and Asia as well as significant portions of the other continents lack the data base, or at least the well-dated base, required to define the 18,000 yr B.P. climate. In the few areas where comparisons were made with the Ice Age climates simulated by general-circulation models, general agreement existed between the geological evidence and the model simulations. Many critical comparisons were thwarted, however, by the lack of model simulations for all seasons at 18,000 yr B.P. Difficulty in validating precipitation anomalies in the tropics also arose because surface-albedo values, which are a vital input to the general circulation models, are estimated from the same evidence that is used to validate the results of the models.