■V22 



THE AMERICAN NATURALIST 



[Vol. XLV 



investigation of inheritance in this form, conducted along bio- 

 metric lines. In their first paper here reviewed they present 

 variation data regarding the four following characters: length 

 of vine, number of pods per vine, length of pod, number of peas 

 per pod, and total peas. The raw data are not given and the dis- 

 cussion is very meager, (iraphs of the variation curves are given, 

 but instead of making these plottings of the actual frequency 

 data as polygons, the authors connect the plotted points by free- 

 hand sweeping curves. This is certainly a simple and expedi- 

 tious, if somewhat naive, method of curve-fitting! It is much to 

 be regretted that such an inadequate, and indeed absolutely 

 incorrect, method of presentation of statistical results should 

 have been resorted to. In the discussion of heredity stress is 

 laid upon the varying degrees of prepotency observed in the 

 transmission of characters by individual plants. To measure 

 this a new "coefficient of heredity" is proposed. The formula 

 for this is 



C = 1/aD, 



where C is the proposed coefficient, o- the standard deviation 

 of offspring and D the difference between the parental character 

 and offspring mean of the same character. It is obvious that 

 the more nearly the offspring are like each other, and like the 

 parent the larger will V, become. It is somewhat unfortunate 

 that this is called a "coefficient of heredity," since this term 

 is in common biometrical usage for a very different constant. 

 Indeed, in their own paper VVaugh and Shaw use this term not 

 only for their proposed constant, but also for the correlation 

 coefficient between parent and offspring. A satisfactory measure 

 of individual (not av<ra<)< ) prepotency is a thing which is badly 

 needed in breeding work. While the constant C proposed by 

 Waugh and Shaw meets some of the conditions which such a 

 measure must fulfill, it unfortunately appears to be of rather 

 restricted significanee and usefulness. The numerical value 

 which it takes for different characters are not comparable one 

 with another. The reason, obviously, is because the numerical 



the relative variability of the character. An elephant and a 

 mouse each equally prepotent with reference to the transmis- 

 sion of any character, say skull breadth, would have very differ- 

 ent values of C for this character. Further, the constant becomes 

 rather difficult to manage in eases of biparental inheritance, or 



