646 



THE AMERICAN NATURALIST 



[Vol. XLVI 



quately described by unvarying hypothetical factors? 

 It is my thesis that if they can not be so described, the 

 Mendelian notation fails. 



Johannsen was the first to show the relative constancy 

 of characters by his beautiful experiments on beans. 

 Since that time, experiments designed to show change, if 

 present, have yielded negative results on bisexual animals 

 such as poultry (Pearl), on plants such as peas (Love), 

 beans (Johannsen 's later work), maize (Shull, East, 

 Emerson), on asexual animals such as hydra (Hanel), 

 paramascium (Jennings) and on asexual plants such as 

 bacteria (Barber and others), and potatoes (East). 



Three critics have appeared. Karl Pearson took up 

 the gage of battle because Johannsen 's work shows the 

 utter untenability of the correlation coefficient as a 

 measure of heredity. He has produced no evidence to 

 uphold his view. Harris, following Pearson for a like 

 reason, has concluded against Johannsen, but has not 

 yet presented his data for public criticism. There re- 

 mains the work of Castle, which he believes is supported 

 by the work of Woltereck. The question to consider then 

 is whether the work of these two investigators justifies 

 the contention. 



Castle states that by selection he has modified a unit 

 character. No one questions that under certain condi- 

 tions changes in characters are made manifest by selec- 

 tion. It has been done again and again. The question 

 as I see it is the following : Are not the facts presented 

 by Castle and the facts of the pure-line workers described 

 most concisely and in a way most helpful to investiga- 

 tion, by the reactions of fixed and unchanging units? If 

 they can not be thus described the use of units is an ab- 

 surdity, for one can not measure or describe by changing 

 standards. 



Castle's principal work on selection is with a fluctua- 

 ting black and white coat pattern— the so-called hooded 

 rat. Tn writing of these experiments, Castle savs (L e* 

 p. 355) : 



