MULTIPLE FACTORS 253 



was tried. The Pi of the reversed minus line be- 

 longed to the 7 (and "7J") generation of the minus 

 selection series. This generation had shown an 

 average grade of -1.56, which represented a re- 

 gression of 0.30 from those extreme individuals 

 (-1.86) of generation 6 from which they had arisen. 

 The range of generation 7 was from to -2.50. 

 Some of the low-grade offspring ranging from 

 -0.37 to -0.87 were chosen for the return selection. 

 They produced 118 offspring whose average was 

 -1.28, a regression of 0.68, which is in the opposite 

 direction from the regression obtained in the 

 former (minus) selection. For six generations the 

 reversed selection went on and carried the race back 

 along its former course, i.e., toward its original 

 condition. The fact that selection in the original 

 direction was still producing some effect when the 

 reversed selection began, means, on the multiple 

 factor hypothesis, that the stock was still heter- 

 ogeneous, in some factors at least, and, therefore, 

 reversing the process would be expected to give 

 the results that Castle and Phillips obtained. 



These important results of Castle and Phillips 

 fulfil so entirely the expectation for multiple factors 

 that they might have been utilized as a good illus- 

 tration of the effects of selection on a group in which 

 a particular character owed its modifications to 

 multiple factors. The results are, however, of addi- 

 tional historical interest because they were used 

 for several years as the favorite experimental '' dem- 

 onstration" for a very different conception. The 



