a number of universals, what determines which
one shall perform the reproduction? Finally, since there are infinite
differences of the universal that might be reproduced, what determines
just which differences shall be reproduced? In this wise the controversy
has gone on ever since the challenge of the old rationalistic logic by
the nominalists launched the issue of empiricism and rationalism. All
the charges which each makes against the other are easily retorted upon
itself. Each side is resistless in attack, but helpless in defense.

In a conception of inference in which both data and hypothesis are
regarded as the tentative, experimental results of the processes of
perception, memory, and constructive imagination engaged in the special
task of removing conflict, ambiguity, and inhibition, and in which these
processes are not conceived as the functions of a private mind nor of an
equally private brain and nervous system, but as functions of
interacting beings,--in such a conception there is no ground for anxiety
concerning the simplicity of data, nor the objectivity of hypotheses.
Simplicity and objectivity do not have to be secured through elaborate
and labored metaphysical construction. The data are simple and the
hypothesis objective in so far as they accomplish the work
where unto they are called--the removal of conflict, ambiguity, and
inhibition in conduct and affection.

In the experimental conception of inference it is clear that the
principles of formal logic must play their role wholly inside the course
of logical operations. They do not apply to relations _between_ these
operations and "reality"; nor to "reality" itself. Formal identity and
non-contradiction signify, in experimental logic, the complete
correlativity of data and hypothesis. They mean that _in_ the logical
procedure data must not be shifted without a corresponding change in the
hypothesis and conversely. The doctrine that "theoretically" there may
be any number of hypotheses for "the same facts" is, when these multiple
hypotheses are anything more than different names or symbols, nothing
less than the very essence of formal contradiction. It doubtless makes
little difference whether a disease be attributed to big or little,
black or red, demons or whether the cause be represented by a, b, or c,
etc. But where data and hypotheses are such as are capable of
verification, i.e., of mutually checking up each other, a change in one
without a corre