438 H. S. JENNINGS AND K. S. LASHLEY 



would be found as a matter of chance; thus, for the first period, so great 

 an excess would occur but 58 times in 10,000 cases. Column 11, giving 

 the odds against so great a deviation as that actually observed, is ob- 

 tained in accordance with rule (5). 



Attention should be directed first to the upper half of the table 

 only; the lower half is designed to meet a possible difficulty, to be 

 taken up later. Table 44 shows that in this experiment, as in 

 that of Miss Cull, analyzed in Part I of this paper, the number of 

 complete pairs among those dead, as well as among those living, 

 is throughout much greater than would be expected if the dis- 

 tribution of deaths had no relation to the pairing. That is, there 

 is a strong tendency for the fate of the two members of a pair to 

 be alike. The condition found is the reverse of that demanded 

 by the theory of sexual differentiation. 



Table 44 shows also, in columns (5) and (9) , that the deviation 

 from the most probable number increases as time passes ; in other 

 words, there is a tendency in the later periods for deaths to occur 

 among the mates of those that have already died out. This 

 tendency we saw likewise in analyzing Miss Cull's experiment. 



The probability of getting the results observed, as a matter of 

 chance distribution, is so excessively small, as shown by columns 

 (10) and (11), that it must be left quite out of consideration. 

 This probability is in fact even much less than that computed 

 for columns (10) and (11), since what we give there is the prob- 

 ability for any deviation of this amount, whether plus or minus. 

 But all the deviations are plus; this decreases the probability 

 greatly. 



It is then absolutely clear that there is something in the pairing 

 which tends to cause the two members of a pair to have the same 

 fate. 



There is one possible source of error that deserves consideration. 

 In removing the pairs to a slide while still united, it is conceivable 

 that both might be injured, in such a way that both would later 

 die. Thus the deaths of a certain number of the complete pairs 

 might be accounted for. 



Wliile there appeared to be no ground for supposing this to be 

 the case in the present experiment, it will be worth while to analyze 



