Gradual, Change, and by Guppy s Method of Differentiation. 611 
to be supposed that the long peduncles are any advantage, or that the 
struggle for existence upon the summit of Nillowekanda is so keen that they 
can have evolved there by infinitesimal variation ? . . . What advantage can 
the two ovules of Polyalthia Moonii and P. pcrsicifolia be against the one 
of the other species? P. rufcscens , another species with two ovules, and 
closely allied to both, occupies the Cochin district of South India, and why 
should there be three species in so similar a country, especially as the Ceylon 
species live in the same district ? And how did the one form get to the 
other, or both arise from a common ancestor, except by mutation ? . . . 
Similar queries might be asked 800 times for the 8co endemics comprised 
in the Ceylon flora. . . . ’ 
The characters that differentiate allied species are as a general rule of 
no importance one way or the other, and cannot have been the subject of 
selection. Only in rare cases do they even allow of intermediate stages. 
The only possible explanation to my mind was that provided by the 
‘ parent and child ’ theory, that parent and child might, and very often did, 
exist side by side. To quote my paper again (p. 14) : ‘ The general principle 
on which India and Ceylon have been peopled with the many species which 
they contain would seem to be that one very common species has spread 
widely, and, so to speak, shed local endemic species at different points, or 
else 1 that one species has spread, changing at almost every point into a local 
endemic species, which has again changed on reaching new localities.’ 
This view of evolution by mutations which in one step—or perhaps but 
less probably in two or more—transform one species by divergent variation 
into another, without in any way necessitating the death or destruction of 
the parent form (so that parent and child might survive together), had been 
published a year previously by Guppy ( 4 ), 2 whose book I had not then 
seen. It was clear that if one much-localized species were derived directly 
from another (usually of wider range) the mutation must almost certainly be 
large, and as there was no reason to suppose a mutation in any given 
direction to be followed by another in the same direction, the change from 
one species to another must probably be due to one or a few mutations, 
each one in the latter case probably changing completely one or more 
characters. Correlation being very common, one would incline to expect 
several or all of the characters of a species to change at once. In Coletis 
elongatus, for example, it was clear that nothing but a big mutation could 
alter calyx or inflorescence, though it was not absolutely necessary that 
both should be altered by the same mutation. 
We have now to go back and consider the work of Guppy ( 4 ), in this 
same direction. His views were also derived from the study of local 
1 Meaning ‘in other cases’. 
2 Priority in the idea of large variation is not claimed for either Dr. Guppy or the writer ; 
it dates at least from Geoffroy Saint-Hilaire ; cf. also the writings of Owen, Mivart, and others. 
