EFFECT OF CONJUGATION 363 



shall come to the same result, finding that in every case the varia- 

 tions are not only relatively, but absolutely, greater among the 

 conjugants. The standard deviations corresponding to the coef- 

 ficients of variation given in table 26 will be found in the tables 

 included in the body of the paper, under the different experiments. 



In just what way is the variation increased in the conjugants? 

 That is, do we find that after conjugation there are more speci- 

 mens with a lower rate of fission, or with a higher rate of 

 fission, or with both? What is the nature of the distribution 

 of the fission rates in each case? 



The fact that the mean rate is lower for the conjugants would 

 cause us to suspect that the increase in variation is at least partly 

 due to a decrease in the rate of fission of some of the lines, while 

 others remain high. Examinations of the data shows that this 

 is largely true. To bring out this point, it will be well to note the 

 comparative range of variation in number of fissions, for the 

 conjugants and non-conjugants, in the various experiments. 

 This is exhibited in table 27. In this table are included the 

 number of fissions for only the lines that lived through the period 

 specified. 



As table 27 shows, at the lower extremity the conjugant lines 

 range much farther than the non-conjugant lines; in every case 

 the lower extreme for the conjugants is below that for the non- 

 conjugants, and in many cases the difference is very considerable. 

 At the other extremity of the range no such great difference is 

 found. The maximum is, as a rule, higher for the non-conjugants, 

 but this is not invariable; in some cases the maximum for the con- 

 jugants is equal to that for the non-conjugants; or even a little 

 greater. 



It appears therefore that conjugation increases the variation 

 mainly toward the lower extremity of the range; it produces many 

 lines whose rate of fission is lower than that for the non-conju- 

 gants, while others remain high. But even in the middle regions 

 of the range, the conjugant lines are less heaped up about the mean 

 than the non-conjugants. These peculiarities may be illustrated 

 by examination of the distribution of the variations in the experi- 

 ments with larger numbers, as given in table 28, for Experiments 



