Deuterium excess (d = δD - 8 * δ18O) values in surface snow are presented for central and east Antarctica. The samples are primarily from Soviet, French, and Australian traverses. The d values exhibit a large change going from coastal sites to high-altitude sites on the ice sheet. The d values are relatively constant at 3 to 6‰ from the coast to an altitude of 2500 m, and at higher elevations d increases steadily to values of 16 to 18‰ at Vostok and Plateau Station. The data is modeled as d versus δD using the kinetic Rayleigh model for isotopes in precipitation developed by Jouzel and Merlivat. The model accounts for kinetic fractionation during evaporation into undersaturated air over the ocean and during snow formation in <−10°C clouds where vapor is supersaturated with respect to snow. The overall pattern of d versus δD can be fit well with a supersaturation function which increases linearly with decreasing temperature and which predicts reasonable values of the supersaturation. Vapor originating from 20° to 60°S was tested with different supersaturation functions. The data could only be fit with moisture originating from 30° to 40°S, indicating that these latitudes are the main source of vapor for snow falling in Antarctica. The conclusion of a mid-latitude vapor source for polar snow agrees with the analysis of d and δ18O seasonal cycles in Greenland snow performed by Johnsen and coworkers. The model was also tested with moisture simultaneously originating from all latitudes from 30°S to the Antarctic coast. The addition of up to 20% of moisture evaporated from latitudes south of 50°, and 5% from latitudes south of 60°, is compatible with low d values occasionally observed in snow near the coast. The conclusion of a “local moisture” effect for coastal and near coastal (<2000 m elevation) snowfall supports a similar conclusion by Saigne and Legrand from their analysis of methanesulphonic acid in Antarctic snow. Finally, the effects of changes in the sea surface temperature and changes in oceanic humidity on the d values observed in Antartic snow are greatly modified during the precipitation process. Hence the interpretation of d values in ice cores should be done in the context of a precipitation model.