VARIATION AND MENDELIAN INHERITANCE 389 



As to the two remaining characters (length of the ear and of 

 the pelvis) the arrangement of the stations follows a quite dif- 

 ferent order in the two cases, while for neither is the order like 

 that for any of the characters previously considered. It must be 

 said, however, that ear length is a rather erratic character in its 

 behavior, some rather perplexing and contradictory results hav- 

 ing been obtained in the course of these studies. The outstand- 

 ing fact here is that the two extremes for ear length are dis- 

 played by the Berkeley and the La Jolla animals, both of which 

 'races' are commonly assigned to the single subspecies gambeli. 

 These differ in average ear length by about If mm., the difference 

 being undoubtedly a real one, characteristic of the mice of the 

 two localities. As to the innominate bone, the difference be- 

 tween the two extremes here shown is doubtless statistically 

 significant, despite its small magnitude. The lesser differences 

 are not so certain. 



Two other modes of portraying these racial differences are 

 shown in figures 3 and 4 and in figure 5, respectively. The 

 former type of chart has been employed by me in many pre- 

 vious papers dealing with variation and heredity in mice. In 

 the case of the present material, I have plotted graphs of this 

 sort for most of the characters here discussed, though only one 

 of these (that for foot length) has been included in this paper. 

 As I have frequently had occasion to explain previously, each 

 of the 'curves' in this figure results from connecting the mean 

 values of this character for the various size-groups into which 

 the individuals of a given race have been divided. Thus, ani- 

 mals of the same size are represented by corresponding points 

 on the various 'curves,' and the comparisons between the races 

 are strictly legitimate. The essential agreement between figures 

 3 and 4 and the scale for foot length in figure 2 is obvious. 



Figure 5 consists of histograms based upon the individual fre- 

 quencies of the various values for two characters in the eight 

 local collections. These, of course, have the advantage of show- 

 ing the total range and variabihty of the respective characters 

 within each race. Such a method of representation would be 

 inappropriate for most of the other characters under discussion. 



