P-, Q-, AND R-TECHNIQUES 205

themselves, nor the qualities of the test stimuli by them-
selves, but the interaction between the two. Which of
the two interacting qualities we treat as constants and
which we treat as variables will depend upon the question
we set out to study and upon the way the experiment has
been planned. This conclusion, which seems more or less
obvious when we are correlating persons, is, I should
maintain, equally true when we are correlating tests. In
either case, the psychologist's factors do not in themselves
describe either persons stimulated or things used to stimulate
them ; they describe the complex relations between the two.

Principle of Reversibility.—To regard the mathematical
analysis as concerned, not so much with the relata, as with
the relations, will confer much the same advantages in
psychology as in physics. It abolishes the need for fixed
distinctions between ' determiner' and * determinate,'
between known * ground * and unknown but predictable
' consequence,3 or (to use more popular terminology)
between c causes ? and c effects.' In experimental psycho-
logy we talk of tests as stimuli which cause reactions ;
in factor-analysis we more commonly talk of abilities as
" causes of individual differences." x I should prefer to
regard each test-process, not as a direct and simple measure-
ment of an isolated trait or an isolated individual, but as a
statement of the point at which an equilibrium is reached
between the external forces (the difficulty of the constituent
test-items) and the internal forces (the mental activities
of the individual tested). Thus, instead of rigidly dis-
tinguishing between variables that are given (cause-factors)
and those which have to be deduced (effect-factors), as we
do in applied or practical psychology, pure psychology
should be content with a set of f equations of condition,'
which all the variables taken together would have to
satisfy. The number of equations would then correspond
to the number of variables that we might require to deduce
or determine, without stating whether the deducible
variables are ' abilities' or future * performances' or
possibly some combination of the two.

Accordingly, if we drop the usual antithesis between
1 E.g. Thurstone, [84], p. 45.