to distinguish between an introspection
(apparent or real) of a feeling or idea, and a process of inference
based on this feeling. The term introspective knowledge must, it is
plain, be confined to what is or appears to be in the mind at the moment
of inspection.

By observing this distinction, we are in a position to mark off an
_illusion_ of introspection from a _fallacy_ of introspection. The
former differs from the latter in the absence of anything like a
conscious process of inference. Thus, if we suppose that the derivation
by Descartes of the fact of the existence of God from his possession of
the idea to be erroneous, such a consciously performed act of reasoning
would constitute a fallacy rather than an illusion of introspection.

We may, then, roughly define an illusion of introspection as an error
involved in the apprehension of the contents of the mind at any moment.
If we mistake the quality or degree of a feeling or the structure of a
complex mass of feeling, or if we confuse what is actually present to
the mind with some inference based on this, we may be said to fall into
an illusion of introspection.

But here the question will certainly be raised: How can we conceive the
mind erring as to the nature of its present contents; and what is to
determine, if not my immediate act of introspection, what is present in
my mind at any moment? Indeed, to raise the possibility of error in
introspection seems to do away with the certainty of presentative
knowledge.

If, however, the reader will recall what was said in an earlier chapter
about the possibility of error in recognizing the quality of a
sense-impression, he will be prepared for a similar possibility here.
What we are accustomed to call a purely presentative cognition is, in
truth, partly representative. A feeling as pure feeling is not known; it
is only known when it is distinguished, as to quality or degree, and so
classed or brought under some representation of a kind or description of
feeling, as acute, painful, and so on. The accurate recognition of an
impression of colour depends, as we have seen, on this process of
classing being correctly performed. Similarly, the recognition of
internal feelings implies the presence of the appropriate or
corresponding class-representation. Accordingly, if it is possible for a
wrong representation to get substituted for the right one, there seems
to be an opening for error.

Any error that would thus arise