A NEW GENE IN DROSOPHILA. 211 



viability may not always produce the same kind of effect. For 

 instance, in culture VIII., the curved class ran far behind all 

 three others, yet in VI. it was the largest class of all. The same 

 relations are shown by the black curved class in XIII. and IV. 

 respectively, and in several other cases. It also appears that 

 the viability difference between two classes differing in two 

 characters is not always merely a summation of the effects pro- 

 duced by these characters separately. Thus in culture VII., 

 since the curved and the black class are each behind the normal 

 class, we might expect the class which is both black and curved 

 to be still further behind yet it is really ahead of both single 

 recessive classes. In culture VIII. black is slightly behind 

 normal, but black curved is far ahead of curved (gray). It is 

 obvious from these considerations that it is not possible to work 

 out "coefficients of viability" and use them for making correc- 

 tions in our data, since with respect to viability the deviations 

 are not constant in amount or direction. However, it is to be 

 noted that when conditions are made as favorable as possible the 

 error from viability is reduced considerably, and often becomes 

 very slight indeed. There is evidence which indicates that dif- 

 ferential viability is often due to unequal sensitiveness to starva- 

 tion, dryness, or similar unfavorable conditions. Several of the 

 cultures recorded in the last table above were purposely kept 

 under various poor conditions (small bottles, little food, etc.,) 

 in order to test this point. Cultures II., III., and VIII. are 

 examples showing the results produced, and III. and VIII. are 

 among the most aberrant cultures in the table. The remedy, 

 then, would seem to be in choosing mutants which are of nearly 

 the same vigor as the normal, and in keeping the cultures in 

 good condition plenty of room and good food. 



Even under these conditions there may be a high mortality, 

 but that this need not always be a differential mortality is indi- 

 cated by an experiment which we have carried out. Three 

 females from the cross of black by curved were tested by back- 

 crossing to black curved males. The eggs were counted daily, 

 and the offspring produced were recorded, with the* following 

 result: 



