WiNSLOW : A STATISTICAL CRITERION FOR BACTERIA 33 



eral distinct characters are compared, it will frequently appear that 

 on the average certain properties are correlated with each other 

 and generally vary together. As mountain tops occur in ranges, 

 so the peaks on the curve of frequency for any single bacterial 

 character are grouped together, by the study of other characters, 

 into larger complexes. Major and minor groups may rightly be 

 considered as constituting genera and species among the micro- 

 organisms. They are not separated by boundary lines but are 

 fixed rather by the centers about which they group themselves. 

 They are genera and species, characterized, not by the description 

 of individual strain's, but by the frequency of occurrence of a con- 

 siderable series of such individuals. 



This method of statistical classification is by no means novel. 

 Ever since Quetelet and Galton began to apply the statistical 

 method to the solution of biological problems it has been recog- 

 nized that species and varieties could be more or less satisfactorily 

 fixed by a study of frequency of distribution. Davenport and 

 Blankinship (Science II. 7: 685) even went so far as to suggest 

 a definite numerical criterion of a species, founded on this basis. 

 According to their suggestion, if a bimodal curve is obtained, the 

 ratio of the ordinate of the lesser mode to the minimum ordinate in 

 the depression between the modes, may be considered as an index 

 of the isolation between the two races. If this ratio is over two, 

 the modes represent species ; if under two, varieties. The numeri- 

 cal criterion fixed by these authors may be criticized as somewhat 

 arbitrary. The general principle is sound, however, and has yielded 

 good results in many fields. Notably in the study of the races of 

 man the statistical method has proved of the highest ser\ace. 

 Curiously enough, among the micro-organisms, where some 

 method of this sort is most urgently needed, statistical study has 

 until very recently been wholly lacking. 



At the Massachusetts Institute of Technology we have been for 

 some years working on the classification of the family Coccaceae, 

 one of the most difficult groups of the bacteria, by this general 

 method. We collected five hundred different strains of cocci from 

 various sources, earth, w^ater, air, and the normal and diseased human 

 body. After some considerable preliminary study eleven charac- 

 ters were fixed upon as having probable systematic importance 



