On a General Theory of Adaptation and Selection. 427 



"... the test of fitness or unfitness has reference to the physio- 

 logical and morphological co-ordination or correlation among the 

 constituents of the whole organism, and . . . any relaxation 

 in either series, in a formative sense or otherwise, results in 

 an instability which may culminate in death, and which expresses 

 itself in structural deviation as well as in a higher degree ot vari- 

 ability." 



It is implied in the foregoing that a distinction may be made 

 between formative correlation and functional correlation. In a 

 later instalment^ the subject is discussed at length, and it is con- 

 tended that the condition of correlation exhibited by the structures 

 of the pupa is dependent upon the correlation of the formative 

 factors or agencies which control the manufacture of the pupa by 

 the larva, while the immediate functional elements are concerned 

 scarcely if at all. The case is therefore quite different from that 

 of the moth, where indeed formative factors of the general condi- 

 tion of correlation must be operative, but where the phvsiological 

 co-ordination of the imaginal structures has a large share in deter- 

 mining the "fitness" of the organism. But entirely aside from 

 the relative values to be assigned to these two classes of factors, 

 the point is that the separate "characters" do not serve directly 

 as adaptive or unadaptive elements of the organism, but they do 

 so only in so far as thev exist in close or loose correlation with 

 other structural and functional characteristics. 



II. 



The truth of the conclusion stated was next put to the test of 

 quantitative determination. The co-efficients of correlation 

 were determined, according to the familiar methods, in the case ot 

 the characters that had been previously treated individually in their 

 relation to elimination, and the co-efficients of correlation of the 

 tw^o groups were compared, %vith positive results. While it is true 

 that the general condition of correlation, regarded as the basis for 

 elimination, is only imperfectly indicated by the degree of correla- 

 tion between any two characters and that the co-efficients of multi- 

 ple correlation involving three or more characters would be more 

 reliable as indices of this condition, yet if the principle be true 



^Now in press, Biometrika. 



