SUMMARY: Of fifty strains of Proteus vulgaris from human infections, seven of the ' vulgaris ' and forty-three of the ' mirabilis ' fermentative type, forty-four were brought into Belyavin's phase A, which represents the modal S form of recently-isolated strains, and analysed serologically by their H and 0 antigens. Three of the six remaining strains were irreversibly R, and three were too unstable antigenically for the routine analysis.The antigens determining the somatic 0 types were resistant to heating at 9 5 O , to ethanol and to dilute HCI, treatments that destroyed the H antigens. Serologically similar 0 antigens were extracted from the bacilli by hot formamide and by ethanol precipitation of tryptic digests.As in other recent surveys of Proteus serology, the classification of strains by major 0 antigens was more clear-cut than by H antigens, though some minor 0 antigens were distributed differently from the major. The forty-four phase-A strains fell into eighteen O-antigenic types, three of these being the classical 0x2, OX19 and OXK types, strains of which were included in the survey. The H antigens of each strain were often apparently multiple, but grouping by major H antigens was broader than by 0 antigens; moreover, H antigens were often common to strains with dissimilar 0 antigens.Apart from the antigenic changes accompanying variation from phase A to phases B or C, there were suggestions of other phasic variations in both H and 0 antigens, which, if substantiated, may necessitate a reconsideration of serological classifications of the group. The three unstable strains in the collection could not be definitely assigned to any O-antigenic type.The serological study of Proteus vulgaris and closely allied microbes covers a period of fifty years. The earlier work was directed, among other things, mainly to finding a species antigen by which P. vulgaris could be identified (Wolf, 1899;Klieneberger, 1908;Cantu, 1911;van Loghem & van Loghem-Pouw, 1912;Horowitz, 1916;Bengston, 1919;Wenner & Rettger, 1919;Taylor, 1928).As a result of these investigations, it was generally conceded that P. vulgaris was antigenically heterogeneous. One type of heterogeneity in Proteus spp., the distinction of flagellar and somatic antigens, was discovered by Weil & Felix who also discovered differences in the 0 antigens of the classical strains, X19, X 2 (Weil & Felix, 1917;Weil, 1920Weil, , 1923Felix, 1923