376 
In type-producing selection, the circumstances which sur 
round all the animals experimented upon are made as 
uniform as possible ; the variations from which selection 
is made are, for the most part, those variations which are 
inevitably associated with sexual reproduction ; and similar 
variants are prevented from interbreeding with individuals 
unlike themselves by a physical barrier. This process is 
continued for several generations until the object which 
the fancier has set before himself has been attained. Now 
let us see how far it is possible that this method should 
obtain in nature. It seems to be probable, if we allow 
the tacit assumptions which the doctrine of Natural Selec 
tion takes for granted. If variations can occur in nature 
which are solely the result of sexual reproduction, if ex 
ternal conditions produce no direct effect upon organisms, 
if necessary birth variations are the sole material for selec 
tion to act on, then it might appear probable that species 
should be transmuted, whether the necessary isolation for 
breeding purposes be effected by an impassable barrier 
or by the survival of the similar favourable variations and 
the destruction of the rest. But in order that species 
should be transmuted in nature by physical separation, 
we must make further assumptions. Similar variations 
must first arise in at least a few individuals of a given 
group ; these must be isolated by a physical barrier, so 
that they breed only with one another. Moreover, in the 
next generation similar favourable variations must again 
arise, apart from the direct or indirect effect of circum 
stances, and these in their turn must be separated. 
But is it likely that this will take place? It is difficult 
to understand how such similar variants should be 
collected together and then isolated by river or moun 
tain, or some natural barrier, in each generation, unless 
indeed they separated themselves, and then no bar