BRISTLE INHERITANCE IN DROSOPHILA 423 



not prove that genetic differences exist between high and low 

 flies. Thus there are two non-genetic factors that may occasion 

 positive correlation coefficients: environment and lack of vigor 

 due to the continued selection of small flies. 



Methods 



In the correlation tables nothing but the progeny of one 

 original pair of flies has been included. As indicated, the 

 return-selected lines probably tend to increase the amount of 

 correlation unduly. However, one of these lines has been 

 included in the tables, namely, the line started from the 16th 

 generation of the high race; two generations of the second re- 

 turn line, started from the 27th generation, have also been in- 

 cluded. The later generations of this last line consisted of 

 such obviously inferior flies that they were omitted. The line 

 of low-grade selection started at the beginning has also been 

 included. The essential genetic difference between this low 

 race and the two return-selected races has been pointed out. 

 There is no suggestion that the low-selected race at the begin- 

 ning had any lack of vigor; the means were immediately influ- 

 enced and there was no question as to the distinctness of the 

 means from those of the high race. On the other hand, the first 

 few generations of the return-selected races showed no influence 

 at all of the selection, but there did gradually appear a slight 

 relative lowering of the means, which may well be due to accu- 

 mulated weakness. The inclusion of these two sorts of low- 

 selected lines will make the comparison of the correlation in the 

 corresponding periods more fair. 



In the tables the families were arranged according to the mean 

 bristle grades of the parents. Since in most cases the parents 

 were of the same, or within one bristle of the same grade, the 

 a,verages were only slightly different from the actual grades. 

 Sons and daughters were tabulated separately. The tables 

 themselves are too numerous for pubhcation in a journal. The 

 correlation coefficients, their probable errors and the ratios of 

 the errors to the coefficients are given in table 1. 



