VARIATION AND MENDELIAN INHERITANCE 373 



The width of the tail stripe (relative, as defined above) appears 

 to be not at all correlated with relative tail length. The mean 

 coefficient actually obtained, which is probably not significant, 

 is slightly negative. This is surprising when we consider that 

 the two characters in question follow much the same order in 

 their degree of manifestation among the various geographic 

 races, and that when these local collections are thrown together 

 and treated as a single population, a fairly high positive corre- 

 lation is found to obtain.^ 



Thus, characters which vary together, when geographic se- 

 quence is considered, may or may not vary together within any 

 single local collection, while, conversely, characters which are 

 correlated within these various local populations may or may not 

 be found to have undergone concomitant modification, when we 

 pass from one loclality to another. 



These relations raise the question whether the interracial dif- 

 ferences in the mean values of various characters belong to the 

 same type as the intraracial or individual differences. By be- 

 longing to the same type I mean having the same sensible prop- 

 erties, and the same mode of hereditary transmission. I think 

 that the ensuing discussion will make clear that the sensible prop- 

 erties are the same, so far as inspection reveals, and that the 

 behavior in heredity is probably likewise the same in the two 

 cases. As to their respective causes, on the other hand, we know 

 too little at present regarding the causes of variation in general 

 to draw any very useful distinctions upon that basis. It would 

 seem obvious, however, that special factors, operating locally, 

 must be responsible for the simultaneous modification of parts 



^ A rather obvious explanation of this apparent contradiction suggests itself 

 here, which, however, I am certain is not the correct one. It might be supposed 

 that an actual positive correlation between these two characters exists within each 

 of the local populations, but that this is too feeble to be appreciable, owing to 

 the limited variability of these populations considered separately. It need only 

 be pointed out that the standard deviations of the local collections, taken singly, 

 are more than half (55 to 80 per cent) as great as those of the mixed assemblage 

 which results when the data are combined. (This has been done for four races 

 only.) Significant positive coefficients of correlation might, therefore, reason- 

 ably be expected within each local race, so far as the extent of variability is con- 

 cerned. 



