291 



THE AMERICAN NATURALIST [Vol. LII 



it does nothing more than reveal the presence of geneti- 

 cally different strains in a mixed population. This, how- 

 ever, is exactly what I wish to do in the present case. 

 The fact that within a "pure line," where the phenomena 

 of heredity should be least obscured, this coefficient is 

 said to be zero is entirely irrelevant to the present situa- 

 tion. What we wish to ascertain is the degree to which, 

 for example, long-tailed parents tend to have long-tailed 

 offspring. Whether these differences among the parents 

 are due to "mutations" or "fluctuating variations," 

 whether they are due to single "unit factors" or "multi- 

 ple factors" or no factors at all, are admittedly matters 

 upon which these coefficients throw no light. Such ques- 

 tions must be decided upon other grounds. 18 



In computing these correlations between parents and 

 offspring, we are restricted to characters which arc in. It- 

 pendent of the absolute size of the individual. 19 Char- 

 acters which fulfil these requirements fairly well are the 

 relative tail length (ratio to body) and the relative width 

 of the tail-stripe (ratio to circumference). My data show 

 that the former is largely, and the latter almost wholly, 

 independent of the size of the mouse. 



The coefficients are given in Table VI. I have not 

 thought it worth while to include their probable errors, 

 since the significance of the set as a whole is indicated by 

 the magnitude of most of the figures and by the fact that 

 all but two out of the 24 are positive. The weighted 

 means of these coefficients, combining the four races and 

 two sexes, are : relative tail length, + 0.297 ; tail-stripe, 



