170 THE VARIATION OF ANIMALS IN NATURE 



characters may be modified in the same species as a result of 

 adaptation to the particular habitat. 



Any attempt to explain such correlation as the expression 

 of single or multiple effects of Natural Selection must, of course, 

 depend on whether Selection is a vera causa. The occurrence 

 of correlation should not be held to be a proof of the action of 

 Selection. 



It is possible in some cases that isolation may make for 

 correlation, as, for instance, when a few individuals of an 

 aberrant form are isolated on an island, so that an association 

 of characters originally accidental is prevented from returning 

 to the normal distribution (by the lack of facilities for crossing 

 with the parent form). Hagedoorn and Hagedoorn (192 1), 

 in particular, have stressed the point that isolation will lead to 

 inbreeding of the isolated stock, with a considerable likelihood 

 of the establishment of a new mean. 



(b) The occurrence of correlation due to the mechanism of 

 heredity has been discussed by Robson (1928, p. 229), who 

 cites a certain number of instances revealed by genetic experi- 

 ment in which, on crossing, character complexes tend to hang 

 together, instead of being dissociated as is the usual fate 

 of independently segregating characters. The majority of the 

 instances are found amongst plants ; but a more limited 

 number occur in animals — e.g. Castle and Wright (19 16), 

 Phillips ( 1 921), Harrison (1916, p. 145 ('segregation en bloc' 

 of specific characters]). The actual basis of such correlation 

 is obscure. The relation between linkage and correlation has 

 been stressed on several occasions, but Robson (I.e. p. 231) 

 makes it clear that it is difficult to attribute the correlation of 

 specific characters to linkage. Sumner (1932, pp. 53-5) had 

 discussed this question in greater detail in connection with 

 his interracial studies of Peromyscus, and finds good grounds for 

 preferring the hypothesis of the multiple effects of single genes. 

 According to Haldane (1932, p. 114), 'a number of cases 

 of multiple action of this kind in Drosophila ' are available. 

 At present very little is known concerning such ' multiple ' 

 effects in animal genetics, and certainly we are not in a 

 position to discuss how far they are influential in producing 

 intraspecific character correlation. 



In any homozygous strain or pure line all phenotypic 

 characters are more or less strongly correlated together until 



