98 CORRELATED VARIATIONS. 



attain a much higher and more permanent value in pro- 

 portion as they are expressed in exact numerical 

 terms. 



To say that some organism or part of an organism is 

 more variable than another is very unsatisfactory, com- 

 pared with the statement that the variabilities of cer- 

 tain characters in the one are of such and such values, 

 and in the other, of certain other values. From such 

 data as these we can compare the variabilities of all the 

 variants exactly, both with each other and with any 

 other variants, and determine what relation, if any, 

 they bear to their systematic importance. We can tell 

 if the variations obey the normal law of error, or if they 

 are asymmetrical in their distribution. In this latter 

 case, we may be able to discover the existence of a 

 tendency to divergence or splitting up of a species, in 

 its very earliest stages. Repetition of our observa- 

 tions at some future period would thus become a sub- 

 ject of especial interest, as we might in such a case hope 

 to detect some change both in the mean values and in 

 the distribution of the variations, indicating that the 

 evolutionary process was still progressing, and the di- 

 rection in which the progress was being made. Fur- 

 ther, as we shall see in a subsequent chapter, by deter- 

 mining the average characters of groups of individuals 

 which have been subjected for some period to the 

 struggle for existence, and comparing them with the 

 characters of other individuals which have not been ex- 

 posed to such a struggle, or by comparing the characters 

 of individuals, which owing to the severity of the strug- 

 gle for existence had been actually eliminated, with the 

 characters of the survivors, we are able to obtain 



