470 WALTER W. MARSHALL AND HERMANN J. MULLER 



CONCLUSION 



The results above recorded give further evidence of the fre- 

 quency with which characters are variable and change geneti- 

 cally without any artificial selection. In the case of the charac- 

 ters dealt with in this paper at least the variation that was 

 observed was certainly not due to contamination or fractiona- 

 tion. This was true in spite of the fact that the factors had been 

 kept in a state of heterozygosis for over fifty generations. 



It should be noted that if the balloon and the other char- 

 acters observed had had a higher survival value than the nor- 

 mal instead of a lower, they might have varied in the opposite 

 direction from that observed and the results might then have 

 been mistakenly attributed to contamination. This fact empha- 

 sizes the importance of not accepting results apparently showing 

 contamination or fractionation of factors at their face value 

 without a thorough factorial analysis. 



