No. 602] SHORTER ARTICLES AND DISCUSSIONS 



103 



ering whether they are also compatible with the alternative 

 hypothesis of a single varying factor. Modification on crossing, 

 decreasing variability, regression, greater variability in F2 than 

 in F^, effective return selection — these are all phenomena to be 

 expected equally on either hypothesis. To cite them is no argu- 

 ment for one hypothesis rather than the other. This point has 

 wholly escaped both Muller and MacDowell, who seem quite un- 

 able to conceive any but the single explanatory principle of mul- 

 tiple factors. 



Putting aside these irrelevant arguments, there remain two 

 points in MacDowell 's discussion which require further consid- 

 eration. They are the same two points which led us in 1914 to 

 hesitate between the alternative interpretations which we con- 

 sidered, but on which we now have fuller evidence. 



But before we go into this new evidence one or two minor 

 points may be noted in which the accuracy of our generalizations 

 is questioned. MacDowell has gone over our 1914 publication 

 in great detail, devoting as many pages to its destructive criti- 

 cism as we to its original exposition, and recalculating the sta- 

 tistical coefficients. It is gratifying to know that he has de- 

 tected in these no serious errors, though his figures differ from 

 ours slightly in some cases. Whether his calculations are more 

 accurate than our twice-checked ones, I am unable to say with- 

 out detailed reexamination of the data. As these are public 

 property, the curious reader may satisfy himself on the point. 

 I do not consider it necessary to go into the matter anew since 

 the substantial correctness of our findings is not challenged. 



MacDowell thinks that we did not sufficiently emphasize the 

 decreasing variability (standard deviation) and the decreasing 

 rate of divergence of the selected races, observed as the selec- 

 tion progressed. These to his mind imply that selection would 

 sooner or later cease to be effective. In this opinion I can not 

 concur, since in neither the plus nor the minus selection series 

 has the standard deviation decreased by half, although sixteen 

 successive selections had been made and the hooded character had 

 been so modified as to be scarcely recognizable longer. Whether 

 one considers the decrease in variability large or small depends 

 principally upon how much importance he attaches to the values 

 found for the first two generations of the experiment, when the 

 numbers of individuals observed were still small and methods 

 of grading them had not yet been fully standardized. Mac- 



