STUDIES ON HIGH AND LOW NON-DISJUNCTION 83 



for instance I crossed exceptional females from the high non-disjunc- 

 tional stock to wild type males. From this again exceptional daugh- 

 ters were crossed to their exceptional brothers, and here we had to 

 expect a low percentage if namely this percentage was dependent on 

 the Y. But in a total of 397 flies I found 17,4 Yo of exceptions. In two 

 further generations exceptional sisters were crossed to exceptional 

 brothers. In the first of them we had to expect a high percentage and 

 in the second one a low percentage. The figures were however in the 

 first generation 23, i % of exceptions in a total of 420, in the second 

 generation 25,2 % in a total of 313. Hence it was obvious that it 

 probably could not l)e the Y-chromosome which was responsible for the 

 high percentage of exceptions in the eosin line of high non-disjunction 

 and it therefore seemed to me to be worth while to study more 

 thoroughly the genetical basis of this high percentage. 



Even from the beginning it must be noted that a rather great 

 difficulty is involved in such an examination because of the fact that 

 there always is a very large range of variation of the exception-percen- 

 tage and that there surely exist a great number of external and internal 

 but not hereditary causes which have an influence on this percentage. 

 Thus although we may use the statistical methods for testing whether 

 two different percentages belong to the same category or not, it is 

 impossible to rely with absolute certainty upon these methods. But 

 on the other hand as we have no other means of interpreting the signi- 

 ficance of the figures from our tables I think that the statistical methods 

 are indispensable. 



I have made use of only one statistical quantity, viz. the mean 

 error. Thus if for instance in one series of experiments there occur n 

 flies with p percent of exceptions and in another m flies with q percent 



of exceptions then the mean error for the first series is+j/^-^^ ^ 



and for the second ±1/ ^ ~^ . Next in order to conclude that the 

 y m 



two values p and q belong to different categories the comparison is 

 made so that the difference between p and q must be more than 3 times 

 1 /fl (100-p) j^ g (100-7) j^^ doubtful cases we may of course also 



be led by the genetic constitution of the culture just tested. To every 

 percent of exceptions that is calculated I have added db the mean error. 

 Since now in nearly all my cultures sex-linked lethals have been 

 present it is not correct to calculate the number of exceptions as 

 percents of the number of hatched flies but it has been necessary to 



