Methods of dissociation were applied to a group of strains of the typhoid bacillus, including recently isolated cultures from “carriers” and several well-known laboratory strains. Typical smooth and rough variants were obtained from both flagellated and non-flagellated forms. Capsules were demonstrated on all forms—rough, smooth, flagellated, and non-flagellated. No serological relationship was observed in agglutination and agglutinin-absorption experiments between the somatic antigens of smooth and rough races of B. typhosus although the existence of a common flagellar antigen was confirmed. Living suspensions of all flagellated and of some non-flagellated rough forms failed to be agglutinated by pure rough somatic antiserum. They did agglutinate, however, when the suspensions were heated to a temperature which destroys flagella and probably also capsules. By intravenous inoculation of rabbits with suspensions of living rough and smooth variants, the relative invasive and toxic character of these forms was determined. No differences in toxicity were observed by this method. Only the smooth variants, however, were found to possess invasive properties as determined by development of the carrier state. Vaccination of rabbits by the intravenous method with adequate dosage of either rough or smooth variants gave protection against approximately fourteen times the minimal invasive dose of smooth living bacilli. The rough vaccines protected distinctly better than the smooth. The sera of rabbits vaccinated with non-flagellated rough vaccines did not possess specific agglutinins for the smooth infecting strain. Those vaccinated with rough flagellated forms possessed only flagellar agglutinins. The inoculation of normal rabbits intravenously with rough variants of the typhoid bacillus resulted in a more marked and persistent leucocytosis than when smooth bacilli were used. Immunized rabbits, regardless of whether rough or smooth vaccines had been used, responded after the usual primary leucopenia with a hyperleucocytosis upon intravenous injection of smooth living cultures, whereas the response of unvaccinated rabbits to the injection of virulent smooth bacilli was relatively slight. The intraperitoneal inoculation of living rough and smooth typhoid bacilli into guinea pigs indicated that the smooth variant was more toxic than the rough. However, the inoculated smooth bacilli multiplied very rapidly in the peritoneum whereas similar doses of the rough variant were disposed of by prompt phagocytosis. The rough variant, as indicated, is phagocyted more vigorously than the smooth in the peritoneal cavity of the guinea pig—both within the peritoneal exudate and on and within the omentum and liver. This phagocytic response to rough organisms in normal guinea pigs parallels the reaction obtained when smooth bacilli are inoculated into the peritoneal cavity of guinea pigs immunized with either rough or smooth vaccines. The results of this experimental investigation are, in every particular, in conformity with the practical experience of the past thirty years in preventive inoculation of human subjects against typhoid fever.