INDUCTION.

As reasoning is from particulars to particulars, and consists simply in
recognising one fact as a mark of another, or a mark of a mark of
another, the only necessary conditions of the exertion of the reasoning
power are senses, to perceive that two facts are conjoined; and
association, as the law by which one of the two facts raises up the idea
of the other. The existence of artificial signs is not a third
_necessary_ condition. It is only, however, the rudest inductions (and
of such even brutes are capable) that can be made without language or
other artificial signs. Without such we could avail ourselves but little
of the experience of others; and (except in cases involving our intenser
sensations or emotions) of none of our own long past experience. It is
only through the medium of such permanent signs that we can register
uniformities; and the existence of uniformities is necessary to justify
an inference, even in a single case, and they can be ascertained once
for all.

General names are not, as some have argued, a mere contrivance to
economise words. For, if there were a name for every individual object,
but no general names, we could not record one uniformity, or the result
of a single comparison. To effect this, all indeed, that are
_indispensable_, are the abstract names of attributes; but, in fact, men
have always given general names to objects also.

CHAPTER IV.

THE REQUISITES OF A PHILOSOPHICAL LANGUAGE, AND THE PRINCIPLES OF
DEFINITION.

Concrete general names (and the meaning of abstract names depends on the
concrete) should have a fixed and knowable connotation. This is easy
enough when, as in the case of new technical names, we choose the
connotation for ourselves; but it is hard when, as generally happens
with names in common use, the same name has been applied to different
objects, from only a vague feeling of resemblance. For, then, after a
time, general propositions are made, in which predicates are applied to
those names; and these propositions make up a loose connotation for the
class name, which, and the abstract at about this same period formed
from it, are consequently never understood by two people, or by the same
person at different times, in the same way. The logician has to fix this
fluctuating connotation, but so that the name may, if possible, still
_denote_ the things of which it is currently affirmed. To effect this
double object (which is called, though impro