Thermomagnetic analysis of natural and synthetic samples containing hematite have confirmed Schwarz's (1969) observations that 1.) extensive reduction of hematite to magnetite occurs when the experiment is performed in vacuum but not when it is performed in air 2.) the rejuction is arrested when the sample is heated in air before the vacuum run. In addition, we have found that the reduction is controlled by back-streaming vacuum pump oils, which buffer the atmosphere in the furnace region, but is not dependent upon reducing agents within the samples. leating of the sample creates a surface layer which is not permeable to the reducing atmosphere and which thus prevents reduction during subsequent heatings. Unusually high Curie temperatures (610¿C) of the reduced phase from synthetic hematite samples may be caused by internal stress in the magnetite which coats the hemite grains.