531 
1921 .] Subspecies and Evolution . 
from such experiments are of little value in studying evolution 
in its natural state. They are, of course, of immense value to 
the economist or poultry-fancier, but the artificial perpetua¬ 
tion of freaks is surely not a state of which Nature approves. 
Natural selection works at the gradual improvement of 
life, and the elimination of what is not good, but has to 
work on certain definite material. Though I realise that 
every branch of life has great possibilities, there are equally 
very definite limitations. For instance, the struggle for 
existence precludes the various branches of life living as 
equals, certain groups always seeking leadership. When 
such accidents as mutational freaks occur, they spring into 
the world on their own responsibility, and are variations 
springing from within, being entirely divorced from environ¬ 
ment. If the change is beneficial or harmless, they are 
allowed to remain and reproduce their freakish variation ; 
if the change is harmful or a handicap, they die. When 
man, however, artificially perpetuates harmful freaks, he 
does so in spite of natural selection. So soon as man with¬ 
draws artificial protection and selection, the freak, thrown 
on the mercy of natural selection, must revert or perish. 
Such is the law for which we have to be thankful. 
Moreover, it seems that nearly all artificially-produced 
races, when removed from artificial conditions, do not retain 
those variations which artificial selection has given them. 
The feral goats which I have seen in Ireland, Scotland, and 
on Round Island in the southern Indian Ocean, the feral 
Pigeon of Mauritius, and the Goldfish which after introduc¬ 
tion to Madagascar devoured the only edible freshwater fish 
in the island, have all tended to revert respectively to wild 
Goat, wild Rock-Pigeon, and the ungainly mud-coloured 
ancestral stock of the Goldfish. 
This leads me to assume that artificial selection does not 
(as Mendelians maintain) alter the gamete, unless artificial 
conditions and selection are maintained for a sufficiently long- 
period to permit cumulative effect. But mutational variation 
under natural conditions remains constant, because natural 
conditions do not appreciably vary. But I admit that the 
