442 EDWIN CARLETON MacDOWELL 



ately upset any selective progress. When the underlying phe- 

 nomena are considered, the possibilities for successful selection 

 become much smaller, because the positive coefficients themselves 

 prove to be very probably due to non-genetic causes. So it 

 appears that the method of the correlation table has not modified 

 the conclusions drawn from the means, namely, that after the 

 early generations in the high-selected race, the bristle-deter- 

 mining germ plasm was uniform and constant; that selection in 

 the early generations produced this uniformity by sorting out 

 differences that existed in the original stock. 



NEW DATA; SELECTION SUSPENDED 



The lack of correlation shown in generations 36 to 49 has 

 been accounted for on the basis that all the matings were made 

 on the same day for each generation, so the environment for 

 each set of families was as near alike as was possible to make it. 

 There may be an objection raised to the conclusion that these 

 uncorrelated generations indicate stable and uniform germ 

 plasm, in that the range of the parents used was very narrow, 

 and further, that full counts were not made of the broods. It 

 may be argued that, even if there is no correlation shown when 

 offspring of parents of high and very little higher grades are com- 

 pared, there might still exist a degree of relationship which 

 would be found if the whole range of parent grades were included; 

 and, further, if the high-grade flies are more apt to appear at 

 the beginning of a culture, the incomplete counts may have 

 hidden a real relationship. To meet these points and to nake a 

 final test of the conclusions already reached, four more genera- 

 tions were raised in large numbers without further selection. 



For a long series of generations the constitution of the germ 

 plasm in the race had been studied, in the attempt to discover 

 the frequency of the changes that might take place therein. 

 Only a limited number of flies could be examined in each gen- 

 eration, and the results were mainly drawn from a comparison 

 of the means in different generations. The correlation method 

 has indicated that similar conclusions are to be drawn when the 

 germinal constitution of the different parents is judged by the 



