Generic Characters in the Coccaceae 149 



application to all bacteria of one fixed series of characters arranged 

 in an arbitrary order tends to suggest a mechanical view of bacterial 

 relationships which is very far from the complex truth. 



In order to obtain a just idea of the real relations of organisms, 

 it is necessary to consider each systematic group by itself. As 

 Robinson has pointed out in an admirable paper on generic classi- 

 fication (Robinson, 1906), "a difference having great classificatory 

 significance in one place may be almost valueless in another." In 

 studying any one group it is therefore necessary to examine afresh 

 each of the various characters used for the identification of bac- 

 teria in general, and to determine its loral value and significance. 

 Secondly, under each character it is necessar}^ to determine how 

 many distinct types of structure or function may occur. This can 

 be done only by measuring the character quantitatively in a large 

 series of individuals, and plotting curves of frequency which will 

 show whether the individual forms fluctuate about one or several 

 modes. This has been attempted by Howe (1904) with good results, 

 for the composition of the gas produced in dextrose broth by organ- 

 isms of the B. coli group. 



Finally, the correlation between various properties should be 

 determined, since it is obvious that the presence of several distinct 

 characters in association is generally of more significance in classi- 

 fication than that of any one alone. 



In the present study we have attempted to obtain the data indi- 

 cated, for certain groups of the Coccaceae. We have measured the 

 easily and definitely measurable, variable characters in 500 sepa- 

 rately isolated races of organisms, and analyzed the data obtained, 

 with two ends in view. We have first plotted the frequency curve 

 for each character to find whether the array varies about one or sev- 

 eral modes, and where the modes are situated, with some measure 

 of the extent of variation about these centers. In the second place, 

 we have calculated correlation factors for the most significant pairs 

 of characters. Each mode on the curves of frequency may fairly 

 be taken to mark a natural species or variety, and the characters 

 which vary together must form the most important basis for the 

 establishment of the larger groups. By such a method alone it is 

 possible to locate those mountain peaks in the chain of bacterial 



