36 WiNSLOw : A statistical criterion for bacteria 



is under consideration. One family or genus may have differ- 

 entiated along the line of carbohydrate fermentation ; another may 

 have divided into groups adapted to a parasitic life on particular 

 tissues of the human body, like the Diplococci ; in a third, obvi- 

 ous differences in reproduction may have arisen, as \\\ the anaerobic 

 spore-formers. This is one reason why the conventional bacterial 

 classifications are inadequate. Migula (System der Bakterien, II., 

 Jena, 1900) and Chester (A Manual of Determinative Bacteri- 

 ology, New York, 1901), for example, recognize only five mor- 

 phological genera among the Coccaceae and divide them all into 

 dichotomous classes by the same arbitrary plus-and-minus tests. 

 Such an arrangement is very far from representing the complex 

 facts of relationship, A difference of great systematic importance 

 in one group may be valueless in another. In each group of the 

 bacteria as many reactions as possible must be examined with an 

 open mind in order to determine which are of systematic impor- 

 tance. In many groups the most helpful properties which mark 

 their characteristic differentiation may be biochemical reactions 

 not yet studied at all. It was the careful examination of fermen- 

 tative power in various carbohydrate media by Gordon and his 

 colleagues that made the classification of the Streptococci possible 

 and it is quite probable that a similar detailed knowledge of the 

 saprophytic Micrococci will be acquired only by the application of 

 tests not now at our disposal. 



The tests chosen for systematic study of the bacteria should, 

 as far as possible, be susceptible of definite quantitative measure- 

 ment. A great deal of the earlier confusion which has surrounded 

 the subject has been due to descriptions of colony-formation, etc., 

 which were expressed in terms wholly dependent on the personal 

 equation of the author. For statistical study, records should, if 

 possible, be capable of expression on a numerical scale. Where 

 this is not feasible, as In comparing pigments of different hues, the 

 observations maybe divided into classes, defined as closely as pos- 

 sible by comparison with known standards. The cultures studied 

 should include a considerable series of the particular group studied ; 

 and it should represent a wide variety of habitats, as the bacteria 

 are usually closely adapted to their particular environment. 



When the observations of several hundred cultures are finallv 



