Various stock strains of B. typhosus were found to produce reacting factors which differed in their neutralizability by anti-stock immune sera. The reacting factors of low neutralizability also possessed low antigenicity. The passage of strains through mice brought about in one instance a lowering of neutralizability of reacting factors and in another instance had no effect. It was possible to increase the antigenicity of the reacting factors by passage of the strain through mice.

Transformation of "stock" strain into "rough" brought about the formation of reacting factors of new specificity. The "rough" reacting factors were neutralized by homologous sera and also differed in their neutralizability.

Reacting factors of various degrees of neutralizability were produced by stock strains which did not display any difference in colony appearance (with one exception). The "rough" reacting factors appeared in strains of typical "rough" colony morphology. However, "rough" strains of the same degree of roughness morphologically differed in the neutralizability of their reacting factors. The changes in neutralizability elicited by passage through mice had no influence on colony morphology.

There seemed to be a parallelism between the degree of agglutinability of stock strains by anti-stock sera and the degree of neutralizability of reacting factors of the same strain.

It was possible to obtain a partial loss of neutralizability by passage through mice with and without change in agglutinability. The agglutination titer of anti-rough serum was not an indication of its anti-rough neutralizing properties.

The degree of neutralizability of "rough" reacting factors paralleled the specific serum agglutinability of the strains producing them.

Abundant precipitation occurred in mixtures of toxic filtrates and immune serum in spite of low neutralization. The filtrates of rough cultures non-neutralizable by anti-stock sera failed to precipitate with the latter. The appearance of neutralizing anti-rough antibodies was coincident with the development of anti-rough precipitates.

The rough strains possessed a lower virulence than the stock strains from which they were derived. However, there was no difference in the virulence of "rough" strains, the reacting factors of which differed in their neutralizability. The filtrates containing "rough" reacting factors were seemingly as lethal to mice and rabbits as were filtrates of stock cultures of B. typhosus.

The possible relation of these findings to the immunological understanding of the course of infectious diseases and to serum therapy is discussed.