5 
variations in fishes, whilst living in their natural state, and distinct 
races have not been formed ; on the other hand, a closely allied 
species—the gold-fish—from being reared in glass or open vessels, 
and from having been carefully attended to by the Chinese, has 
yielded many races.”—(The Variation of Animals and Plants under 
Domestication, vol. ii.,fi. 236.) 
Without selection individual differences will be swallowed 
up in the average of the race, through the principle which 
Mr. Francis Galton calls the regression to mediocrity. 
This principle differs from the ordinary action of atavism : 
which, as I have pointed out, will sometimes cause an 
individual to resemble a remote ancestor, and which pro 
duces an influence not to be easily calculated. The 
regression to mediocrity, on the other hand, represents the 
effect of the ancestry as a whole ; and Mr. Galton seems to 
have succeeded in ascertaining the exact amount of this 
influence. This principle, I venture to believe, only applies 
to those variations which are inevitably and necessarily 
associated with sexual reproduction. Thus understood, it 
involves the absolute necessity for the strictest isolation 
of similar variants, but at the same time, it cannot be 
understood to mean that this regression to mediocrity will 
take place when similar variants are isolated for breeding 
purposes, for such an assertion would be diametrically 
opposed to the experiences of the cattle breeder and the 
pigeon fancier. 
“ It will be seen” (says Mr. Galton) “from the large values of the 
ratios of regression, how speedily all peculiarities that are possessed 
by any single individual to an exceptional extent, and which blend 
freely together with those of his or her spouse, tend to disappear. A 
breed of exceptional animals, rigorously selected and carefully 
isolated from admixture with others of the same race, would become 
shattered by even a brief period of opportunity to marry freely.”— 
{Nattire. vol. xxxiii.,f. 297.) 
But Mr. Galton emphatically states that the law of 
regression does not invalidate the principle of selection :—