376 FRANCIS B. SUMNER 



might be held to account for the parallel modification af parts 

 which ordinarily do not vary together. 



Now, I regard it as highly probable that such racial differ- 

 ences as relate to pigmentation have been produced in some way 

 by environmental agencies. And our problem would doubtless be 

 simplified if we could regard the other differences named as hav- 

 ing arisen simultaneously through this same set of environmental 

 factors. Against this supposition, however, is to be set the fact 

 that these various characters are not always modified in the 

 same direction throughout the entire range of the species. For 

 example, the desert mice have narrower tail stripes and less pig- 

 mentation generally than those of Berkeley, whereas the mean 

 tail length is almost identical in the two races. Of course, it is 

 possible to rejoin that the effects of environmental influences in 

 any given case are probably very complex, and that while one set 

 of conditions might call forth parallel modifications in two differ- 

 ent characters, another set of conditions might call forth diver- 

 gent ones. In making such assumptions, we should, of course, 

 be venturing upon very uncertain ground, but I cannot conceive 

 of an explanation of the curious relations here considered, with- 

 out some sort of appeal to local (i. e. environmental) factors. 



I have already stated that when we throw together certain of 

 the local collections and treat them as a single population, a 

 decided positive correlation appears between two characters (tail 

 length and tail stripe) ^ which were not correlated within the local 

 collections taken separately. This, indeed, is a mathematical 

 necessity when two groups of individuals, differing in the mean 

 values of two characters, are mixed together. The characters in 

 question are inevitably found to be correlated in the mixed popu- 

 lation. It might seem, at first, to be equally inevitable that the 

 gradual migration and dispersal of these mice would bring about 

 a similar intermixture, resulting in a measurable correlation be- 

 tween two such characters. That this has not actually resulted 

 is doubtless owing to the slowness of the process of dispersal. 

 If these animals were continually traveling great distances and 



' This would doubtless be true of certain other pairs of characters, for which 

 my data are not yet complete. 



