454 H. S. JENNINGS AND K. S. LASHLEY 



for a, 17 for b; et cetera. When it is recalled that all the lines 

 were propagated with the most extreme precautions for keeping 

 them uniform; that the progeny of a and b were cultivated in 

 separate dishes, handled separately, and their records separately 

 kept, to be compared only at the end, it will be evident that there 

 is a most striking degree of resemblance, due to intrinsic causes, 

 between the progeny of the two members of pairs. 



This comes out strongly when we ask the question whether the 

 differences between the two members of the pairs are greater or 

 less than the differences between two individuals compared at 

 random. This is readily worked out from table 49 by the method 

 described on page 425. The average difference in number of fis- 

 sions for 24 days between the two members (a and 6) of the pairs 

 (comparing each a with each b of its own pair) is found to be 2.349 

 fissions. The average difference when each individual is com- 

 pared with every other is 8.590 fissions. Thus the average differ- 

 ence between the members of actual pairs is less than one-third 

 as great as that between individuals taken at random. The 

 pairing has increased the resemblance between the progeny of the 

 two members of the pairs in a high degree. 



WTien, on the other hand, we turn to the split pairs we find a 

 complete contrast in these respects. To begin with, we are 

 confronted by the fact, brought out in my previous paper ('13, 

 page 351), that there is no differentiation in rate of fission among 

 the different lines of propagation (or at least extremely little). 

 Thus there is little or no opportunity for any special degree of 

 resemblance between the two members ot and b, of the split pair. 

 But proceeding to examine the matter directly in the second half 

 of table 49, we find indeed that no special similarity between the 

 fissions for the a and b of split pairs is seen. The contrast in 

 this respect with the conditions found in the pairs is most striking. 

 If we determine the average difference between the two members 

 of the split pairs we find it to be not less that that between individ- 

 uals taken up at random. Indeed, for the particular case of 

 table 49, it turns out that the difference between the members of 

 split pairs is slightly greater than for random comparisons, the 



