To learn more about liquid metal distillation as it might be applied to nuclear fuel recovery and reprocessing, liquid metal distillations involving the tin-zinc, cadmium-bismuth, and indium-zinc binary systems were conducted in an evacuated chamber. An x-ray fluorescence spectrometer provided a continuous chemical analysis of the distilling surface during each run. This information was used to evaluate the validity of various theories and assumptions concerning surface depletion, oxide contamination, and turbulence effects. The existence of a large surface depletion effect in nonturbulent metal distillations was proven. However, the level of turbulence necessary to eliminate concentration gradients was found to be much lower than that assumed by some designers of commercial equipment. The presence of surface oxides was often an important factor in determining the enrichment and rate of distillation. The Langmuir-Knudsen theory was shown to be unreliable when liquid diffusion or surface oxide resistances were significant. A more complete approach involving the principles of transport phenomena was developed. An analytical solution was derived for the nonturbulent case and was tested using the spectrometer data.