338 H. S. JENNINGS 



for the conjugant and non-con jugant lines, in table 21. The 

 constants are given, not only for each of the five periods of table 

 33, but also for certain of these periods taken together; likewise 

 for the first week (column 1). 



Table 21 shows that: (1) in every case the mean rate is higher 

 in the non-conjugants; (2) in every case the standard deviation 

 (measure of the absolute amount of variation) is greater in the 

 conjugants; (3) in every case the coefficient of variation (measure 

 of the variation relative to the mean) is much greater in the 

 conjugants. 



Since these measures are based on the number of fissions for 

 long periods under identical conditions, they can hardly be held 

 to represent meaningless accidental fluctuations, but rather actual 

 differentiations. They show further that these differentiations 

 are much greater in those that conjugated during the last epidemic 

 than in those that did not. This conjugation therefore caused 

 inherited differentiations within the pure line. Whether the 

 fewer inherited differentiations among those that did not con- 

 jugate during the last epidemic are due to previous conjugations 

 we cannot tell in this case, but must refer the reader to the account 

 of Experiment 15. 



The inheritance of the rate of fission in these cases may be 

 demonstrated, for those that prefer this method, by working out 

 the coefficients of correlation. The numbers we are dealing with 

 are of course small, but significant, owing to the great number of 

 generations dealt with. We may take the fissions during the 

 first two periods of table 33 and by determining their correlations 

 with the fissions in the same lines for the second two periods, get 

 a numerical expression of the inheritance. For the conjugant 

 lines we find that the coefficient of correlation thus taken is 

 0.5031 ± 0.1346. For the non-conjugant lines it is 0.5627 ± 

 0.1331. 



(A full treatment of the inheritance of the fission rate^ by 

 biometric methods, with adequate numbers, will be given in 

 another connection; together with an analysis of the relation of 

 this method of measuring inheritance to other ways of dealing 

 with the matter.) 



