BRISTLE INHERITANCE IN DROSOPHILA 137 



Standard deviations 



In the return race the similarity of the standard deviations 

 with those of the high race was cited as evidence of the sim'- 

 larity of the two Hnes. In this extracted race the standard 

 deviations argue in like manner that the high race is different 

 (fig. 10 A and B). The standard deviations have been omitted 

 for groups 11, 12 and 13 since there were such a few individuals 

 included in these groups. The standard deviations of the ex- 

 tracted race are lower than those of the high race. There seems 

 to be something that hinders the appearance of the high bristle 

 grades. It is not environment. One can see from the means 

 that favorable conditions affect the two races in the same way, 

 since both races rise at the same time; yet the rise in the ex- 

 tracted race is never as great. The standard deviations support 

 the conclusion that the extracted race does not vary so much 

 as the high race, that there is a genetic difference between these 

 two races. 



Conclusions 



The establishment of a low race from the high selected race 

 by means of a cross, stands in strong contrast with the unsuc- 

 cessful attempt to produce a low race by selecting low grade 

 parents directly from the high race. As in the high race, the 

 first few generations of selecting were successful, and the curves 

 of the offspring and parents are parallel; the lower parents pro- 

 Fig. 9 Means of the extracted race which was started from the extra flies 

 extracted from a cross between the high race and normal wild. Parents and 

 offspring are compared with parents and offspring in the uncrossed high race. 

 The data are grouped according to the half month in which the matings were 

 made. The last nine averages shown in each curve include all the flies from 

 one generation; at this time all the matings for one generation in each race were 

 made on the same day, so the similarity of the food in these last nine generations 

 results in a greater similarity in the environment of the contrasted races, and 

 there appears a closer parallelism in the means. Parents weighted as in pre- 

 ceeding curves; 'A' based on the sons, 'B,' on the daughters. The first selection 

 immediately separated a race that was distinct from the high race. After the 

 first few generations, fluctuations in the parental means do not occasion cor- 

 responding changes in the means of the offspring. 



