72S 



THE AMERICAN NATURALIST [Vol. L 



8. As Muller has correctly reasoned, successful return 

 selections would be expected on the multiple factor view, 

 supposing the races were still " heterozygous even after 

 generations of selection" (Muller, :14, p. 571). It may 

 be added (MaeDowell, :15, p. 95) that the failure of a 

 return selection to reduce the average, as long as the ad- 

 vance selection was progressing, would be strong evi- 

 dence against the multiple factor interpretation. As 

 long as there remained any heterozygosity in the race, 

 both advance and peturn selection should succeed in mov- 

 ing the averages. 



9. In no case should return selection lower the averages 

 at a rate faster than advance selection was raising them 

 at the same time. That is, a certain degree of heterozy- 

 gosity will permit a certain rate of advance or decline of 

 the averages. The return selections that were started 

 after several advance selections did not show a decline 

 that could be compared with the sudden advance that oc- 

 curred in the first selected generation. The plus selec- 

 tions had reduced the heterozygosity and had thus set 

 closer limits on the effectiveness of return selection, as 

 well as of further advance selection. So return selections 

 from later generations should be less effective than re- 

 turn selections from the earlier generations. In the plus 

 race, generation 7, there is a difference of .84 between the 

 average of the offspring of rats selected to continue the 

 plus race and the average of the offspring of rats chosen 

 for a return selection. In generation 12 the correspond- 

 ing difference is .60. In the minus race the difference 

 between the averages of offspring from high and low 

 grade parents in 7, 8 and 9 average .50. In generation 12 

 the difference is .28. The numbers of rats are very small 

 in most cases, but it is interesting to note that as far as 

 they go they seem to show that return selection is less 

 effective in the later generations than in the earlier ones. 



10. Castle has shown that the increase in variability 

 in the first generation of a cross between the plus and 

 minus races may be considered an indication of segrega- 



