COAT-COLOR IX HYBRIDS OF PEROMYSCUS EREMICUS. 



53 



based upon 12 pairs of variables, particularly since the characters 

 in queston are not wholly independent of one another, but are 

 themselves in some instances, rather strongly correlated. On the 

 other hand, it hardly seems likely that this entire correspondence 

 is accidental. 1 



rding to the theory of "multiple factors," we should, in 



'ral. expect the increase of hybrid variability to be greater, in 



proportion a- tin- parental difference was greater. Consideration 



of the sjinpl, --ible case, involving a single pair of factors, 



> be true. Thus, the positive correlations here re- 



port.-,! ^ranting their significance are in harmony with this in- 



n. 



To surn up, then, the results of our present inquiries: (i) We 

 find, for a number of measurable characters, affecting the pig- 

 itation of the hair and skin, a blending type of inheritance, 

 both in the F, and F, generations, following a subspecific cross. 



The variability of these characters is, with a single excep- 

 tion, somewhat greater in the second hybrid generation than in 

 the first. 



, ) This increase in variability is greatest for the length of 

 the Mihurminal yellow band, least of all (slightly negative) for 

 th<- " en ratio." The former character is the one which 



differ- most widely in the two parent races, and is likewise most 

 ni;l\ ! iry. as judged by parent-offspring correlations. 



I !'< latt( r differs but slightly in the two parent races, while the 

 individual differences within a race do not appear to be inherited 

 at all. With one conspicuous exception, there is a rough pro- 

 portionality between the amount of the difference between the 

 parent ; n respect to a given character, and the amount of the 



inn : variability in the F, generation. 



14) In general, the coefficients of correlation between F t par- 

 ents and their F. 2 offspring are lower than those between the pure- 

 nice parent- and their F, offspring, despite certain elements in our 

 procedure which would tend to increase correlation in the former 



1 A Minilnr calculation, based upon all of the characters of all of the Pero- 

 myscus hybrids thus far considered (76 pairs of variables) yielded a coeffi- 

 cient of + o.ji>4 + .074. One of the series, the " Carlotta-Calistoga " lot, 

 yielded, IMNM-MT, a IK-;: alive coefficient ( 0.17). 



