EFFECT OF CONJUGATION 337 



probable, or perhaps certain, that such differentiation arises in 

 consequence of conjugation. But it leaves unsettled the ques- 

 tion whether such inherited differentiations may not arise also 

 in other ways. 



To give clear results o*n this point, the experiment should have 

 been performed as follows: A single individual of k should have 

 been isolated, allowed to multiply by fission; watched continu- 

 ously till the first conjugation occurred, then the experiment 

 should have been performed with these conjugants and non- 

 conjugants. If inherited differentiation appeared among the 

 non-conjugants in such a case it could not be held to be due to 

 conjugation. 



These conditions are fulfilled in Experiment 15 on another 

 race, to be described. They were likewise fulfilled in the latter 

 part of the present experiment, and repeatedly in experiments on 

 race k in 1912. But unfortunately race k has lost its power to 

 flourish in slide cultures; in every case with the later experiments 

 on this race all the lines have died out after a few weeks of cul- 

 ture. It would be of interest to carry out the experiments with 

 race k, in view of its history of eight repeated self-fertilizations, 

 and efforts will be made to find a successful method of slide cul- 

 ture for it. In the meantime the results of Experiment 15, with 

 race E, give clear results on the main questions at issue. 



The results of the present experiment therefore leave open the 

 possibility that heritable differentiations may arise in other ways 

 than by conjugation. Do they furnish positive evidence that 

 heritable differentiation actually does arise as a result of conju- 

 gation? As we have seen, all our many experiments show that 

 conjugation increases the variation in rate of fission between the 

 lines. This is true (as already set forth) for the first week of the 

 present experiment. Furthermore, if we compare the \'ariability 

 of the conjugant and non-conjugant lines of table 33,- we find again 

 thaf the conjugants are much more variable. We are of course 

 not here dealing with random samples, but since both sets were 

 selected to give as much variation as possible, a comparison of the 

 variations may be of significance. The means, standard devia- 

 tions and coefficients of variation for various periods are given 



