348 H. S. JENNINGS 



for September 4 and 6 (that is, + 2 + 2 + 1 + 3 + 3 = 11); for the 

 second half, 8 by, September 8 to 16 ( = 17); and similarly for all analo- 

 gous cases. Thus in the correlation tables a given preceding period 

 may sometimes be counted twice, since it gives rise to two lines of prog- 

 eny, and is therefore correlated with both. This of course introduces 

 no error into the coefficient of correlation. The other constants (given 

 in table 23) were computed with each period counted but once. 



Results. Comparative examination of tables 34 and 35 shows 

 a very great difference between the progeny of the pairs, and 

 those of the split pairs, in respect to variation and differentiation 

 in the rate of fission. This is well shown by observing the range 

 of variation in the two cases. In the split pairs the slowest lines 

 show in twenty-one days 18 fissions, the fastest, 28 fissions. 

 In the pairs the range is (for twenty-four days), from 10 to 35. 

 If we reduce these latter numbers by one-eighth, in order that 

 thay may compare directly with those for the split pairs, the 

 range becomes for the pairs 8.75 to 30.6, as compared with 18 

 to 28 for the split pairs. From September 8 to 16 the range for 

 those derived from the pairs is 1 to 17; for those from the split 

 pairs, 8 to 15. 



Working out the mean number of fissions, with the standard 

 deviation and the coefficient of variation for the pairs and split 

 pairs during a number of different periods, we obtain the results 

 shown in table 23. Here the data are given for the first and 

 second halves of the experiment; also for the entire period. In 

 order to have certain periods which are absolutely identical in 

 every respect for the two sets, I give also for the pairs the data 

 for the ten days extending from September 6 to September 16, 

 this period coinciding with the 'second half for split pairs. 



As the last columns of table 23 show, the variability in fission 

 of the lines descended from conjugants was, for the entire period, 

 four times as great as that for those descended from non-conju- 

 gants. In the first half of the time it was about twice as great; 

 in the second half five times as great. 



Examining tables 34 and 35 to discover the cause of this great 

 difference in variation, we find that the descendants of those that 

 have conjugated are differentiated into a number of distinct 

 lines, with different rates of fission. This will at once be evident 



