ON SUB-FAMILIES. 991 
arrangement, the greater is the probability of our 
discovering the order of nature, it becomes essential. 
to ascertain how far the laws regarding the com- 
bination of sub-genera into genera will assist us to 
combine genera into sub-families. We have sup- 
posed in the latter case, that the naturalist has found, 
in all his perfect genera, the prevalence of a deter- 
minate number of minor divisions; he is now, 
therefore, to try the strength of the law thence as- 
sumed, upon a more extensive scale. First, we must 
combine our genera in such a way that four, five, 
seven (or whatever the assumed number may be), 
make a circle of their own, more or less complete. 
We shall then have a certain number of circular 
groups, forming one of larger dimensions ; and, pro- 
ceeding in this way to form other assemblages of 
the same kind or rank, compare their respective 
contents. The first test of every such circle will be 
that its primary divisions or genera are also circular : 
the second, that these divisions or lesser circles, in 
regard to their nwmber, are definite. If we find 
that their average number is greater or less in the 
sub-families, than in the genera, we must conclude 
one of two things; either that the number of types 
vary in different groups according to the rank or 
value of such groups, or that we have not yet dis- 
covered what is the true number most prevalent in 
nature. Every principle of sound reasoning is 
against the first of these suppositions; for if we 
suppose that natural groups are perfectly independ- 
ent of any definite number of divisions, then (setting 
aside all experience to the contrary) we virtually 
deny uniformity of design in the details of nature, 
