THE EAW MATERIALS OF EVOLUTION 103 



measuxements of 2,600 ip.en, taken at random, 

 Wallace notes that there is 1 of 4 ft. 8 in. and 

 1 of 6 ft. 8 in. ; 12 of 5 ft. and about 12 of 6 ft. 4 in. ; 

 equal numbers at equal distances from the mean 

 of 5 ft. 8 in. 



This tedious work of registering the variations 

 that occur may lead us right into the heart of 

 the matter; thus the asymmetry or skewness of 

 the curve may show us at a glance that the species 

 is moving in a definite direction as regards the 

 particular character measured, or the formation 

 of a double-humped curve may vividly bring 

 home the fact that the species is dividing into 

 two sub-species. Thus, by a statistical path, we 

 are brought face to face with the most vital of 

 all facts — revolution creatrice. 



(3) Correlation of Variations. — Another im- 

 pression which we get from some of the modern 

 work on variation is, that the living creature 



of the alignment of the rifle. But as these factors act with equal 

 frequency in every direction, it follows that the point of thickest 

 distribution of the shots will still remain at the centre of the target. 

 Now, variation is found to follow precisely the same law. If measure- 

 ments of some character are taken in a large number of individuals, 

 it is found that there is a mean measurement in the neighbourhood 

 of which the individuals are most thickly clustered, and that the 

 further the distance from the mean, the fewer are the individuals 

 represented. The analogy goes yet farther: for, just as in the 

 case of the good and bad riflemen, we found the shots to be in 

 close juxtaposition or more widely scattered, so in the case of 

 variation, it is found that the divergences from the mean are in 

 some cases far more accentuated than in other cases ; that is to 

 say, the degree of constancy or variation in different organs is 

 very different. But in all cases the variation can be represented 

 by a geometrical curve, the ordinates of which are proportional 

 to the terms in the expansion of the binomial (a + h). Occasionally 

 the individuals are foimd to cluster round two or more points of 

 thickest distribution, and it is then inferred that they belong to 

 two or more differeat races." — Edinburgh Review, Jan. 1909. 



