VARIATION AND MENDELIAN INHERITANCE 395 



Two characters which are correlated with a high degree of cer- 

 tainty are relative tail length and the number of caudal verte- 

 brae. It seems, on first thought, curious that relatively longer 

 tails should tend to have a larger number of vertebrae, while the 

 absolute length of this appendage should play no appreciable 

 part in the matter, I have not tested directly the correlation 

 between absolute tail length and the number of caudal verte- 

 brae, but the fact that the longer animals of my series (having, 

 as a consequence, longei" tails) do not have more vertebrae than 

 the shorter ones, renders improbable the existence of such a cor- 

 relation. It must be pointed out here that the slight differences 

 met with in the number of the caudal vertebrae have little part 

 in determining the differences in tail length, whether between 

 races or individuals. These depend mainly upon the size, 

 rather than the number of the separate bones. 



In determining correlations between the various other pairs of 

 characters, a different procedure has been adopted, owing to the 

 fact that these characters are all strongly correlated with body 

 length, and therefore, in a population of mixed size, necessarily are 

 correlated with one another. For this, as well as for other pur- 

 poses, I have divided up the animals of each local collection into 

 groups of individuals differing by less than 2 mm. Correlations 

 have been determined for each size-group containing ten or more 

 individuals, and the means of these coefficients employed.^i 



Positive coefficients of probable significance have been ob- 

 tained for tail and foot and for tail and skull. The correlation 

 between tail and ear length is far less certain, while none appears 

 to exist between the tail and the pelvis. On the other hand, 

 the foot and the pelvis seem to be correlated with a consider- 

 able degree of probability. It is worth adding that the existence 

 of a positive correlation between tail length and foot length, as 

 well as between foot length and that of the pelvis, is made yet 

 more probable from an inspection of certain series of mice which 

 were measured before my methods were fully standardized, and 

 which have therefore not been included in the present compu- 

 tations. 



21 The various methods here used have been discussed in a previous paper 

 (Journal of Experimental Zoology, April, 1915). 



