722 



THE AMERICAN NATURALIST [Vol.XLIX 



the average. With the rats, however, a very different 

 condition exists. The average is not changed by increase 

 of high-grade individuals merely or chiefly. At the pres- 

 ent time every individual in the plus selection series and 

 nearly every individual in the minus selection series is of 

 higher grade (plus or minus respectively) than any indi- 

 vidual in the race at the outset. It is not a fallacious 

 change of averages which has taken place ; a genuine and 

 permanent racial change has occurred, following step by 

 step upon repeated selection. Generation by generation 

 new grades of offspring have come into existence, more 

 extreme in character than any which existed before, and 

 simultaneously with the advance of the outer limit of vari- 

 ation the inner limit has receded. No great change in 

 variability has attended the selection. The standard devi- 

 ation has decreased somewhat to about three fifths of its 

 original amount, but has scarcely altered in the last eight 

 or ten generations ( see Tables I and II ) . Rather there has 

 occurred a change in the modal condition of the character, 

 about which fluctuation continues very much as before. 

 When the position of the mode changes, as a result of 

 selection, the position of the average and of the upper and 

 lower limits of variation change with it. In a word the 

 character changes. 



In our 1914 publication Phillips and I were conservative 

 about asserting a change in the single Mendelian unit- 

 character manifestly involved in the hooded pattern. We 

 suggested the possibility that other as yet undiscovered 

 factors might be responsible for the apparent changes 

 observed and awaited the result of experiments then in 

 progress to show whether such a possibility was admis- 

 sible. I have no hesitation now in saying that it is not. 

 All the evidence we have thus far obtained indicates that 

 outside modifiers will not account for the changes ob- 

 served in the hooded pattern, itself a clear Mendelian 

 unit. We are forced to conclude that this unit itself 

 changes under repeated selection in the direction of the 

 selection; sometimes abruptly, as in the case of our "mu- 



