448 H. S. JENNINGS AND K. S. LASHLEY 



lines die out of 478 the most probable number of pairs is but 

 1. The actual number is five times as great as the most probable 

 number, so that it is clear that among those dying during these 

 four days something tends to cause both members to die if one 

 dies. Common injury to the two in separating them would have 

 this effect. 



Compare this with the relations in the next six days of the first 

 ten-day period. There are left at the beginning of this six days 

 218 complete pairs (436 lines); of these 29 died during the six 

 days and these included 2 pairs. Thus the number of deaths is 

 greater than in the first four days, but the number of pairs in- 

 cluded is less than half as great. By formula (1) we find that when 

 there die 29 out of 436, the most probable number of pairs is 1, so 

 that here we have an excess of 1 pair. The probability for 1 pair 

 is 0.3942; for 2 pairs, 0.1811, so that 2 pairs would occur, as a 

 matter of chance, about half as often as 1 . The deviation from 

 the most probable number is 1, and by rule (5) we find that prob- 

 ability of so great a deviation is 0.60582. Hence it is more prob- 

 able that we should find so great a deviation as this than that 

 we should not, even if the distribution of deaths is quite independ- 

 ent of the mating. There is thus no positive evidence here that 

 the two members of ike split pairs resemble each other more than 

 any two individuals taken at random. 



In the last sixteen days of the experiment (following the first 

 four-day period), 70 of the lines died out, including 8 pairs. By 

 formula (1) or (2), the most probable number of pairs when 70 

 lines die, out of 436, is 6, so that we have here a deviation of 2 

 pairs. By rule (5) we find that the probability of so great a 

 deviation is 0.3038, so that we should find such a deviation in 

 about one case out of three, though the distribution of deaths be 

 quite independent of the mating. Such a result cannot be con- 

 sidered to give any positive evidence that the two members of 

 the split pairs tend to have the same fate. 



We may also examine the results for the entire twenty days, 

 comparing it with the results for the first four days. Of the en- 

 tire 478 lines, 99 died out during the twenty days, including 15 

 pairs. The probable number of pairs is 10, so that we have an 



