1 62 THE VARIATION OF ANIMALS IN NATURE 



on the causes very far, nor did he attempt to distinguish 

 the various phenomena to which the name ' correlation ' is 

 given. 



The distinctions made by Diirken and Kerr can be harmo- 

 nised, if we realise that Diirken's ' relation ' and ' correlation ' 

 are causal types of association and correspond to Kerr's ' physio- 

 logical correlation ' ; while Diirken's ' combination ' includes 

 Kerr's ' gametic ' correlation as well as other phenomena. 

 Thus we can include in it (a) character associations produced 

 by the mechanism of heredity in its distribution of segregating 

 characters (e.g. effects of linkage, strains homozygous for 

 several characters, etc.), and (b) equally fortuitous association 

 produced by the coincident effects of external causes operating 

 simultaneously on the individual. 



It is desirable, before proceeding further, to obtain some 

 general idea as to the extent to which the characters distin- 

 guishing species and races, etc., are correlated. Were such a 

 measure obtainable, it would give us an idea as to the extent 

 to which these groups are homogeneous for their diagnostic 

 characters. Taxonomic experience, of course, prepares us 

 for the result that the degree of correlation is very varied, 

 probably on the whole rather low. The value of the available 

 data is rather dubious, as what we obviously want to know 

 about is the correlation of hereditary characters, and in sys- 

 tematic data little attention is paid to the discrimination of 

 fluctuational from hereditary characters. 



A great deal of statistical information is available as to the 

 correlation of miscellaneous characters, but very little con- 

 cerning those which distinguish groups. The exact analysis 

 of the variation — e.g. of pairs of related species or races — from 

 this point of view has been very little studied, and more work 

 of this kind is desirable. The facts we give are slight in amount, 

 but we believe they may be typical of a larger array. It must 

 be borne in mind that such studies as are available are made 

 on limited sections of populations, and we have no means of 

 saying how far the correlations indicated are characteristic 

 of the groups over their entire range. Lastly there is available, 

 as far as we know, no analysis of all the diagnostic features of 

 a pair of allied species. 



We will first give (a) some data concerning the correlation 

 of characters within species, and then (b) examples of the 



