126 EDWIN CARLETON MACDOWELL 



that was carried on for 8 generations before it was lost. A defi- 

 nite limit had been set above which no flies should be used as 

 parents. It happened there were such a few flies available be- 

 low this limit that the number of matings was small. Since 

 there is a tendency for the low bristle grades to appear on small 

 flies, the selection of low grades accomplished the sorting out 

 of the smallest and weakest flies as progenitors of the race. 

 These two explanations seem to account for the fact that no 

 flies at all appeared in the matings made from the 8th genera- 

 tion of this line. A second series of return selections was 

 started from the 26th generation of the high race, but this series 

 also came to a premature termination. In summarizing these 

 data, the flies have been grouped according to the time their 

 parents were mated, instead of according to the generation. 

 This has been done so that, in studying different selected hues, 

 the flies compared would have more nearly similar environ- 

 mental conditions. It frequently happened that the matings 

 for one generation would be made during several weeks so that 

 families in different generations would be hatching at the same 

 time. This would often result in famihes in different genera- 

 tions being raised under more similar conditions than families 

 from the first and last matings in a single generation. All the 

 progeny of flies mated in the same half month have been grouped 

 together and corresponding groupings have been made of the 

 families in the high selected race. These classes are called 

 'groups.' All the families mated in the first half of June, 1915, 

 are averaged and plotted in the first point of the curve in figure 

 7; all the matings made in the same time in the high race are 

 averaged as the first point in the curve so marked. There are 

 also shown in figure 7 the averages of the parents selected (dotted 

 lines) that produced the offspring averaged in the solid lines. 

 The data are given in table 2 A and B. The averages of the 

 parents have been weighted according to the numbers of their 

 sons in 'A' and according to the numbers of their daughters in 

 'B.' Although the low and the high selected parents average 

 about four bristles apart, the averages or their offspring are 

 very close together. In both the return and high selected races 



