388 



SCIENCE 



[N. S. Vol. XXIX. No. 740 



sent natural families or genera. Some of 

 these groups are based on morphological dis- 

 tinctions while others are simply held together 

 by certain physiological resemblances. And 

 in practically all cases as soon as the firm 

 ground of morphological characters is left, 

 and attempts are made to make use of physi- 

 ological differences, we find systematic bac- 

 teriology becoming simply determinative 

 bacteriology, and all semblance of natural re- 

 lationships' is lost in a confusion most be- 

 wildering. 



It has remained for the work of the Wins- 

 lows to bring order out of chaos, to show us 

 how it is possible to delimit the different 

 groups of bacteria and to determine their 

 natural relationships, with just as sure a 

 footing, whether we make use of morphologi- 

 cal or physiological characters. 



Their method of defining bacterial groups 

 is by a study of the numerical frequency of 

 various characters in a large series of cul- 

 tures. It matters not whether the characters 

 are morphological or physiological as long as 

 they are meas\irable. It is true this method 

 of defining species is not original with these 

 authors; anthropologists and students of 

 variation and heredity have developed it for 

 the study of their particular facts. Even 

 among bacteriologists it was being used at the 

 same time that the work of the Winslows was 

 going on by Andrewes and Horder in England 

 for the classification of the streptococci. But 

 it was our present authors who pointed out 

 the importance of this method for work with 

 the bacteria, and it is to them that all credit 

 should be given for working out the method 

 and applying it on a large scale to the problem 

 of bacterial classification. 



It is not necessary here to refer to the 

 method of biometry. It depends on the fact 

 that fluctuating variations, when measured in 

 a considerable number of individuals, group 

 themselves in a curve which follows the 

 simple mathematical law of chance. If two 

 large arrays of individuals are measured the 

 curves obtained are practically identical. But 

 if arrays from different origins are measured 

 the shape of the curves will differ, as well as 

 the position and height of the modes. Such 



curves measure the peculiarities of a group as 

 a whole, and serve to discriminate the differ- 

 ent types, even though particular members of 

 the groups are indistinguishable. By extend- 

 ing the observations to include the correlation 

 of characters in the different racial types, the 

 statistical method will indicate the systematic 

 relationship of the different types. 



As the authors themselves say: 



The biometric methods, which have proved so 

 useful in the study of the races of man, promise 

 to be of even greater value in the systematic 

 analysis of types among the bacteria, where so 

 many factors combine to preserve varietal differ- 

 ences on so wide a scale. If individual strains 

 only are considered, an infinite series of differences 

 appear. If the same strains are considered statis- 

 tically, that is, if the frequency of a given char- 

 acter be taken into account, it is apparent that 

 certain combinations of characters are much more 

 common than others. Measurement of almost any 

 character by quantitative methods shows that the 

 bacteria examined group themselves on a simple 

 or complex curve of frequency. The modes of this 

 curve indicate centers of variation about which 

 the individuals fluctuate; and these centers of 

 variation are the real systematic units of the 

 group. The recognition of such centers, as specific 

 types, offers the natural and satisfactory com- 

 promise between systematic multiplicity and vague 

 generalization. The grouping of specific types is 

 an even more important problem than the defini- 

 tion of the types themselves; and here the correla- 

 tion data obtained by biometric study are of 

 assistance. A true natural classification is tree- 

 like and includes branches and twigs of varying 

 grades of importance. Genera of bacteria should 

 be aggregates of those specific types which are 

 most nearly related; and the basis of the relation- 

 ship will differ in each individual case. . . . 

 Finally, the results may be analyzed with two 

 ends in view. First, each center of numerical 

 frequency, marking a group of organisms varying 

 about a distinct type in regard to a single definite 

 property, may be recognized as a species. Second, 

 those species which are bound together by the 

 possession of a number of similar properties may 

 be constituted as genera, and larger groups of 

 genera, still characterized by some characters in 

 common, may receive the rank of families or sub- 

 families. 



The recognition of these principles will 

 throw a flood of light upon all our future at- 



