No. 546] INCONSTANCY OF UNIT-CHARACTERS 355 



corner grocery than in the U. S. Bureau of Standards, 

 and the dishonest tradesman will select effectively for 

 diminished size among the various quart measures 

 offered on the market, unless his selection is carefully 

 restrained by legislation. Similarly urinal unit char- 

 acters are modifiable under selection; only one blindly 

 devoted to a contrary theory will be able long to shut his 

 eyes to this fact. For several years 1 have been engaged 

 in attempts to modify unit-characters of various sorts 

 by selection and in every case I have met with success. 



I shall speak first of the case least open to objection 

 from the genotype point of view which requires: 



1. That no cross breeding shall attend or shortly pre- 

 cede the selection experiment, lest modifying units may 

 unconsciously have been introduced, and 



2. That only a single unit-character shall be involved 

 in the experiment. 



These requirements are met by a variety of hooded rat 

 which shows a particular black and white coat-pattern. 

 This pattern has been found to behave as a simple Men- 

 delian unit-character alternative to the self condition of 

 all black or of wild gray rats, by the independent investi- 

 gations of Doncaster and MacCurdy and myself. The 

 pigmentation however in the most carefully selected race 

 fluctuates in extent precisely as it does in Holstein or in 

 Dutch Belted cattle. Selection has now been made by 

 Dr. John C. Phillips and myself through 12 successive 

 generations without a single out-cross. In one series se- 

 lection has been made for an increase in the extent of the 

 pigmented areas ; in the other series the attempt has been 

 made to decrease the pigmented areas. The result is that 

 the average pigmentation in one series has steadily in- 

 creased, in the other it has steadily decreased. The de- 

 tails of the experiment can not be here presented, but it 

 may be pointed out (1) that with each selection the 

 amount of regression has grown less, i. e., the effects of 

 selection have become more permanent; (2) that advance 

 in the upper limit of variation has been attended by a 

 like recession of the lower limit ; the total range of varia- 



