Gradual , Change , and by Guppy s Method of Differentiation. 627 
species by ‘ large * differences that suddenly appear. Neither of the extremes 
can now command general acceptance, and it is suggested that they should, 
so to speak, exchange partners. 
Natural Selection is essentially a theory of progressive adaptation, and 
it is shown that adaptation usually involves correlation, and there is no 
reason to suppose that gradual stages, always correlated, can be picked out, 
whilst often the one half of the selection could not be made till the other 
half was complete, and the result of the latter only would be fatal, as 
in tropical lianas, for example, where the twining stem could not begin to be 
selected till the weak and flexible stem had been produced, though the 
latter, with no means of support, would be at a hopeless disadvantage. 
A brief account is given of my former investigations of the endemic 
species of Ritigala, and it is shown that the chief distinguishing characters 
of Coleus eloiigatus (to take a definite example) do not allow of intermediate 
forms, nor can they be the subject of Natural Selection. The only reasonable 
explanation is that they were formed by large changes. And the same is 
the case for a vast proportion of important distinguishingcharacters between 
species generally. 
Guppy’s Differentiation theory, which is upon much the same lines, is 
then touched upon, with some notes by Dr. Guppy himself. 
Other published work is then described in brief, all pointing to the same 
general conclusion, and it is shown that destruction of intermediates, usually 
called in upon the Darwinian theory, will not explain the facts, nor, in the 
majority of cases, is there any reason for it. 
It is then pointed out that nothing but experience can decide what are 
the most useful distinguishing characters in any given case. To make up 
one’s mind beforehand to use chiefly characters A, B, and C, for example, 
would only be to court disaster. 
Instances are given of characters sometimes of great importance as family 
diagnostics and in other cases only generic or specific ; and the characters of 
the ‘abnormal ’ members of Rubiaceae are given in full detail to show how 
in a single large family no character whatever need necessarily be steadfast 
throughout the family. The incidence of rumination of endosperm, some¬ 
times (e. g. in the palms) a generic or even only a specific character, is 
described in more detail, but such detailed work must in general be left for 
later papers. 
Finally, the very conclusive evidence in favour of large mutations that 
is given by the work of Mr. Yule and the author is touched upon. If genera 
follow in their evolution the rules of compound interest, then it is practically 
impossible to suppose them formed in any other way than by sudden muta¬ 
tions. All the evidence that is produced in this paper points to the same 
conclusion. 
