1. Altogether 117 strains of Brucella, belonging to different types and isolated from different parts of the world, have been examined by the thermoagglutination, salt agglutination, acid agglutination, and specific serum agglutination tests.

2. The results obtained by the thermo-agglutination and the serum agglutination tests are in close agreement; there is a fairly high degree of correlation between these tests and the acid agglutination test, and a rather lower correlation with the salt agglutination test.

3. Generally speaking, a strain which is highly thermo-agglutinable is frequently agglutinated by salt, is usually agglutinated strongly by acid, and reacts to a paramelitensis, but not to an abortus serum.

4. A strain which is moderately thermo-agglutinable is seldom agglutinated by salt, is frequently agglutinated by acid, and reacts either with an abortus or a paramelitensis serum, or with both sera.

5. A strain which is not thermo-agglutinable is not agglutinated by salt, seldom reacts markedly to acid agglutination, and is generally agglutinated by an abortus, but not by a paramelitensis serum.

6. There remain, however, a certain number of strains, particularly of the porcine and bovine abortus types which, though non-thermo-agglutinable, inagglutinable by salt, and reacting only with an abortus serum, yet show some degree of acid agglutination.

7. Of the twelve porcine strains examined only one strain was strongly thermo-agglutinable; of the forty-seven bovine strains only two were strongly thermo-agglutinable, a further two showing a milder degree of thermoagglutinability; of the forty-seven melitensis strains eight were strongly, and thirteen were moderately thermo-agglutinable; while of the eleven paramelitensis strains ten were strongly thermo-agglutinable.

8. These results are taken to indicate, in accordance with the suggestion made by certain previous workers, that those strains which are non-thermoagglutinable, are not agglutinated by salt, and are agglutinated by an abortus but not by a paramelitensis serum, represent the smooth form, while those strains which are strongly thermo-agglutinable, are frequently agglutinated by salt, and are agglutinated by a paramelitensis but not by an abortus serum, represent the rough form.

9. If this interpretation is correct it will be noticed that the great majority of the porcine and bovine strains examined were of the smooth type, that nearly half the melitensis strains were partially or completely rough, while all but one of the paramelitensis strains were rough.

10. Whether melitensis strains have a greater tendency than abortus strains to undergo the smooth to rough transformation it is difficult to say with certainty, but the reports in the literature and the observations in the present paper render this probable.

11. By serial passage through broth at 5-day intervals, it is possible to transform smooth strains of all three types into the rough form. This transformation appears to occur more readily and to proceed further in a given time with melitensis than with abortus strains; but since only three strains of each type were examined, the results may have been determined as much by chance selection of strains as by any greater inherent tendency of the strains of the melitensis type to undergo variation.

12. It is clear that none of the tests employed suffices to differentiate individual strains of abortus and melitensis. The thermo-agglutination test and the agglutination test with specific smooth and rough sera do, however, enable a differentiation to be made between smooth and rough strains of all types.

13. In the present paper no attempt has been made to distinguish abortus and melitensis strains by specific agglutination and absorption tests. The general failure of workers hitherto to obtain any clear-cut serological distinction between these types may possibly be due to the fact that many of the strains with which they worked were either partially or completely rough. Since the rough antigen seems to be more or less alike in strains of all types, it is clear that its presence would tend to obscure any difference that might exist between the smooth antigens of the different types. If such a difference does exist it is probable that it will be elicited only by a comparison of absolutely smooth strains.