INTRODUCTION. xvii 
gorous state becomes R. Doniana of Woods, which is 
the intermediate state between it and Sabini. This 
passes into villosa through R. gracilis of Woods. By 
a partial return to its original appearance, that is to 
say, by again losing its hairs, but retaining its glands, 
R. involuta becomes R. myriacanthas Another branch 
from spinosissima is through rubella which passes into 
alpina by the dwarf alpine variety of the latter. A 
vigorous sort of spinosissima in two or three generations 
might produce R. hibernica, and this may be traced 
into Sabint without much difficulty. From R. Sabini 
having lost its setae proceeds R. tomentosa, whose va- 
riety mollis brings them together; and we are ac- 
quainted with every gradation from tomentosa to ca- 
nind. 
Before we proceed to consider how far these cir- 
cumstances may be allowed to affect the arrangement 
of the genus, it may be worth while to consider what 
is, or ought to be, understood by a species. Cuvier 
tells us it is ‘the union of individuals descended from 
each other, or from common parents, and of those 
which resemble them as much as they resemble each 
other.” (Regne Animal 1. 19.) De Candolle defines a 
species to be “ the assemblage of all the individuals 
which resemble each other more than they resemble 
others; which can, by reciprocal foecundation, pro- 
duce fertile individuals ; and which reproduce them- 
selves, by generation, in such a manner that they may 
all by analogy be supposed to have descended originally 
from a single individual.” (Theorie ed. 2. 193.) Now 
if these definitions, which are purely hypothetical, be 
the test by which a species is to be tried before it is ad- 
mitted as such, it results, from the illustrations I have 
c 
