234 THE FACTORS OF THE MIND

less to commit myself, openly or tacitly, to any particular
solution. Pure psychology is concerned solely with in-
quiries on the intermediate level, below the loftier plane of
the metaphysician, who alone is authorized to talk of
realities and to include or exclude souls and their trans-
cendental powers, but above the humbler plane of applied
psychology, where the everyday parlance of individual
substances and persons, of cause and effect, of abilities and
will, again becomes permissible. In pure psychology, as in
pure physics, the word * causes' ought in every rigorous
argument to be carefully eschewed, and be replaced by some
non-committal description such as that of ' functional
relations.' The psychologist, as it seems to me, must accept
the warning which Eddington has addressed to the physicist,
namely, that " so-called causal events are to be thought of
merely as conspicuous foci from which the links radiate/'
in short, as the centres of naturally cohering sets ; the
c links * are not " lines of force or power," but simply
" determinate laws or mathematical equations connecting
the events " ; and it is " with these links or relations only,
and not with the relata, that the humble scientist is con-
cerned."

Relativity in Psychology.—The very terms * co-relation '
and c co-variance' ought to have suggested at the outset
that we were about to analyse relations between qualities
and variations in qualities (that is, differences between
qualities), and were no longer dealing with qualities them-
selves. No one has ever thought of the variance or the
standard deviation of a group as measuring a mental ability
or power : and hence, I imagine, no one would regard the
covariance of a group as measuring a mental ability or
power. Now, by its formula, a coefficient of correlation is
nothing but a covariance, arbitrarily standardized. Un-
fortunately, however, in many psychological textbooks a
coefficient of correlation is described as a measurement of a
tendency. As a result, the reader insensibly substitutes
the idea of a single measurable thing, the c tendency,'
for the idea of a { constant' (and a very inconstant constant)
summarizing a system of measurements ; and the student is
apt to believe that a correlation coefficient describes some-