BRISTLE INHERITANCE IN DROSOPHILA 



435 



Extra bristles 



If the means of the high and the return-selected race differ on 

 account of non-genetic causes, the positive correlation coefficients 

 obtained when the two lines are correlated at the same time are 

 due to these same non-genetic causes. For it can be shown 

 that the high-selected line within itself at this time did not 

 produce higher-grade offspring from higher-grade parents. In 

 figure 3 and table 3 are given the correlation coefficients for a 

 period covering the time of the return-selected race. 

 In order to group together flies with more similar environments, 

 the families are arranged according to the time of mating (in 

 half -month periods) instead of by generations. Groups 1 to 10 

 correspond in time to the return-selected line. The sons do 

 not show a single group with significant positive correlation; the 

 daughters show just one such group. So it is evident that the 

 significant positive correlation shown in generations 18 to 23 in 

 figure 1 does not mean that the parents in these generations of 

 the high-selected race differ genetically. This conclusion holds 

 whether the non-genetic interpretation of the lowered means in 

 the return-selected race is accepted or not. 



Generations 24, 25, and 26 include but one family each. 

 The breeding conditions were very unfavorable at this time. To 

 save the race all the flies available in generation 26 were mated, 

 however low their bristle grades. Soon after this the bottles 

 were transferred to a constant-temperature room (90°F.); the 

 size and bristle number of the flies at once rose higher than at 

 any other time (MacDowell, '17, fig. 3, p. 114). Unfortu- 

 nately, all the matings for one generation were not then made 

 at the same time, so different generations entered the warm 

 room at the same time. The families in the second half of this 

 middle period, when grouped by generations show four positive 

 and two negative correlation coefficients. When arranged by 

 half-month periods, and the return selections omitted, as shown 



