Fractions of bovine white matter, prepared by the methods of Folch and Lees, were studied for chemical composition and for their ability to produce experimental allergic encephalomyelitis in rabbits. Evaluation of the disease and of the lesions in animals injected with the more active fractions at several dose levels permitted comparison of the antigenic activity of these materials. When tissue was fractionated by the methods of Folch and Lees, antigenic activity was found in the chloroform-methanol extract but not in the denatured tissue residue. This activity was traced to proteolipides A and B and to the lower phase, more specifically the ether-soluble fraction of the lower phase. Proteolipide C was inactive. Correlation of the chemistry of fractions with their antigenic activity suggested two possibilities: (a) that there might be two antigens, one proteolipide and the other non-proteolipide; or (b) that a small specific proteolipide is responsible for all the observed activity. The high concentration of acetal phosphatide in the ether-soluble lower phase suggested that compounds of this type might be the hypothetical non-proteolipide antigen, but this hypothesis was disproved by analytic study of active and inactive materials. The possibility that proteolipide might account for all the antigenic activity was strongly supported by the experimental finding that total lipide and proteolipide progressively lost activity as proteolipide was degraded by adequate processing. The use of an entirely different method for preparing total lipides free of proteolipide (the colloidal iron technique) indicated that this loss of activity did not result from incidental removal of some non-proteolipide antigen. These tentative conclusions are in agreement with those of Tal and Olitsky and provide a satisfactory interpretation of the findings of Goldstein et al. The very fact, however, that the suggested proteolipide antigen would amount to no more than 1 per cent of the total chloroform-methanol extractives leaves open the possibility that some unrecognized trace substance may be the antigen. Skin tests with the various fractions indicated some cross-reactivity between proteolipides A and B and the ether-soluble lower phase fraction and a fair correlation of positive skin reactions with disease. This finding is compatible with the suggestion that the same antigen is present in both of these types of material. When the disease produced by whole tissue or fractions was evaluated by the use of the proportion of animals developing disease, the day of onset, and the severity of the histologic lesions, it was found that fractions produced milder disease of later onset than intact tissue at all dose levels. The disease-producing activity was not enhanced by increasing the dose; i.e., it appeared to reach an asymptotic maximum below that obtainable with whole fresh tissue. This finding suggests both a quantitative loss of activity and a qualitative change during the initial chloroform-methanol extraction, a procedure which denatures all proteins in the tissue residue. A comparable change appeared to occur in whole white matter stored at -15 degrees C. for 15 months and thawed and refrozen several times during this period. The later fractionation steps resulted in no apparent loss of antigenic activity. A scoring method employing the same type of data to estimate the actual relative antigen contents of different preparations is presented in the Appendix.