303 
whole adaptive morphology of nature. How immense, thei'efore, 
must be the number of unuseful variations. These are probably 
many hundred of times more numerous than the useful variations, 
although they are all, as it were, stillborn or allowed to die out imme 
diately by intercrossing.”—{Journal 'of Linnean Society. Zoology, 
vol. xix., p. 348.) 
If this view is the correct one, Natural Selection secures 
the survival of a few favourable variations and the des 
truction of a few unfavourable ones, but does not prevent 
a large number of unuseful variations from surviving; 
but it is assumed that these will be stillborn or die out 
immediately by intercrossing. Here, then, the principle of 
regression to mediocrity does more than Natural Selection 
to maintain the efficient stability of species. 
But it is not at all so clear that these unuseful varia 
tions will be stillborn and will be immediately absorbed 
in the average of the race ; and, if not, there is no reason 
why they should not intermarry with the more typically 
perfect ; and in this way the average—not the ideal 
perfection—of the race would be maintained. 
If, now, we take the hypothesis that the variants range 
themselves in a central column, on either side of which 
there are struggling variants, then the principle of regres 
sion to mediocrity will do all that is required ; the plus 
and the minus will neutralise one another, and Natural 
Selection is not required. 
Next, let us suppose that the variants range themselves 
in two columns on either side of a central line. In the 
first place, we must be well assured that these variants 
are real sheep and goats—all useful on the one side and 
all prejudicial on the other side. Only on this hypothesis 
could Natural Selection act in the way supposed ; for if 
there were a large output of unuseful variations, Natural 
Selection could not touch them. But if Natural Selec 
tion did choose all the column to the right because it