394 FRANCIS B. SUMNER 



absolute values which have been determined are strongly and 

 positively correlated with body length. ^^ The correlation is 

 highest in respect to the three skeletal characters which are 

 included. Of the two relative characters, the ratio of tail to 

 body is seen to be negatively correlated with body length. 

 That is, larger mice have proportionally slightly shorter tails. 

 On the other hand, the relative width of the tail stripe (ratio to 

 (circumference of tail) does not appear to be correlated signifi- 

 cantly with the general size of the animal. 



Much more instructive from our point of view are the correla- 

 tions of the various characters, other than body length, with one 

 another (table 3). That between tail stripe and relative tail 

 length has been computed in the entire undivided populations, 

 irrespective of size. Of the sixteen different figures (the races 

 and sexes being treated separately), it will be seen that six are 

 positive and ten negative, the mean for the entire series being 

 slightly negative. Since, however, the probable errors for 

 these single coefficients are, for the most part, nearly or quite 

 as great as the coefficient themselves, it is quite unlikely that the 

 preponderance of the negative values is significant. We may 

 fairly assume, therefore, that the two characters are not appre- 

 ciably correlated. 



I have likewise thrown together (though not in the present 

 table), the data from the four 'races' taken north of San Fran- 

 cisco Bay, and treated the entire lot as a single population, the 

 sexes, however, being dealt with separately. Deviations from 

 the grand averages were employed in the computations. The 

 stations represented are Calistoga, Duncan Mills, Fort Bragg, 

 and Carlotta.2° Fairly high positive coefficients now appear 

 between the two characters last mentioned (+0.366 for the 230 

 males and +0.351 for the 176 females)' The significance of 

 this fact has been discussed in the preliminary section of the 

 present paper. 



1^ Exception should be made of the number of caudal vertebrae. 



2" Since the Carlotta and Eureka collections are nearly identical in their 

 mean characters, I have not included the latter. The four sets used were trapped 

 and measured during the same season and are possibly more nearly comparable, 

 on this account, than ones taken in different years. 



