682 The American Naturalist. [August, 
of very similar individuals not connected by intermediate individuals 
with other groups different in details of structure, form or color. An 
apparent and probably sufficient cause for this is the close similarity 
of all Diplopoda in life-histories, habits and food. All are scavengers, 
able to subsist upon a variety of decaying vegetable, or even animal 
matter, and there has been scarcely any response to calls for special 
adaptations to life as parasites, commensals, or under other changed 
conditions. The species of Diplopoda are not only extremely local in 
distribution, but are generally confined to almost identical habitats, 
removed from which they do not long survive. 
Supposing the Diplopoda to be a natural group descended from a 
common ancestor, we are compelled to believe that such differences as 
appear among them are the result of accumulated variation not greatly 
influenced by external selective causes. Hence, existing differences 
indicate in general much more remote developmental divergence than 
in groups which have entered more thoroughly into the struggle 
for existence by responding to the demands of varied conditions. In 
this respect the Diplopoda offer a most striking contrast to the Hexa- 
poda, and the results are in accordance; there are more millions of 
species of Hexapoda than there are thousands of Diplopoda. __ 
Having accepted a criterion of species, the classification into higher 
groups is perhaps largely a matter of convenience ; but convenience, 
scientific accuracy, and the recognition of affinities, alike demand con- 
stant attention to the fact, that the value of any character depends 
primarily upon its constancy, not upon the apparent degree of diver- 
gence, This is merely the reiteration of the chief axiom of systematic 
science, but the abundance of systems which completely ignore this 
fundamental idea are evidence that much reiteration is still desirable. 
While in some natural groups it seems necessary to recognize sub- 
divisions not definable by any constant character or complex of char- 
acters in the Diplopoda, we may conveniently proceed upon somewhat 
better ground, and require that the genera and larger divisions shall be 
limited by definite structural characters. 
A dichotomous classification is theoretically the only exact one, for 
the reason that three or more natural groups could never be ex 
to be separated by exactly equivalent structural differences. Practi- 
cally, however, a dichotomous system is inconvenient by reason of 
the great number of categories necessary in properly recognizing 
affinities. Hence, it is not a valid objection to the usual or multi- 
fid form of classification that the natural divisions arranged under 
the same category are not of the same rank, that is, not remote from 
