286 FIRST PRINCIPLES OF NATURAL CLASSIFICATION. 
stant ; but the corallines assume an endless diversity of 
form, although the general structure of the species is 
essentially the same. Much experience, therefore, is 
sometimes necessary to discriminate a species from a 
variety: in general, however, a variety may be defined 
as local or accidental, whose peculiarities are not per- 
petuated in the next generation, and which cannot be 
traced in more than a few individuals. It must be 
‘again mentioned, nevertheless, that these observations 
are applied only to animals in a state of nature ; since 
it is well known that the greatest variation of form, 
colour, and even of structure, have been produced by 
long domestication. 
(347.) Having now sufficiently developed all those 
principles of the system of nature with which we are 
as yet acquainted, it follows that no arrangement of her 
groups yet discovered can be natural, unless they exhibit 
these principles in their details. It has frequently been 
observed, and with great truth, that “‘a natural arrange- 
ment will stand any test.” But the test itself must 
first of all be proved genuine. It is not a sufficient 
test of our groups, that the individuals composing them 
are placed in a circular series ; because hundreds of such 
circles can be made out, the fallacy of which, did no 
other test exist, can never be discovered. Neither is a 
group sufficiently verified by making out its parallel 
relations of analogy with another group; because, as al/ 
contain the same denomination of types, we may happen 
to compare a family with a genus, and, finding that 
both have parallel analogies, may be led to fancy that 
both are of equal value: both groups, indeed, may pos- 
sibly be natural; but if we merely confine our analysis 
to these, without investigating others which are con- 
terminous, we may combine them falsely, and thus 
throw a whole order into confusion. Parallel relations 
must also be of a definite character, or the imagination 
may be led astray: hence the necessity of verifying every 
group, not only by the system of sepresentation, but 
also by the law of variation and succession of the pri- 
