374 FRANCIS B. SUMNER 



which do not ordinarily vary together. I shall return to this 

 point presently. 



It must be borne in mind that all of the differences which I have 

 dealt with between the local races relate to the average condition, 

 and do not hold constantly for every individual of the groups 

 under comparison. In fact, a large proportion of the individuals 

 belonging to two adjacent groups might be placed indifferently 

 in either, without the transfer being detectable by any test known 

 to me. It is only the most widely separated collections, e.g., 

 those from Humboldt Bay and the Mohave Desert, which differ 

 so much in respect to certain characters that the frequency poly- 

 gons for these last do not overlap more or less broadly. Between 

 such a condition of distinctness as this and one of practically 

 complete identity we have many gradations. In some cases it 

 is only by the comparison of probable errors that we are enabled 

 to say whether or not two collections differ significantly in 

 respect to a given character. 



It must be remembered, too, that the local 'races' with which 

 I am dealing are highly artificial groups. My collections are 

 simply samples, taken at various arbitrarily chosen points from 

 a perfectly continuous population. Whether or not these local 

 differences in the average or modal condition would be completely 

 bridged by collections taken at stations sufficiently close to one 

 another remains problematic. It would seem almost inevitable, 

 however, that interbreeding would lead to such a complete con- 

 tinuity, at least in the absence of some sort of geographic bar- 

 riers. For mice of even the most widely separated of these races 

 appear to be fully fertile inter se. 



It appeared early in the course of these studies that the various 

 racial differences were hereditary. The races 'bred true,' so far 

 as could be detected by the methods employed and allowing for 

 certain abnormalities of form to which all of the races were sub- 

 ject when reared in captivity. 



I have also shown that the variations within each local race — 

 or some, at least, among them — are rather strongly hereditary. 

 As stated in an earlier paper, the parental-filial correlation in 

 respect to tail stripe and to relative tail length averages, in each 



