134 THE FACTORS OF THE MIND

So far, we have confined our attention to the measure-
ment of a single, empirical variable only. Granted that
we can find a scale for comparing differences in the same
characteristic, what scale can be devised for comparing
differences in different characteristics ? That, after all, is
the essential prerequisite of correlation. Evidently, if we
are to combine measurements of two or more variables, other
assumptions must be added to our list—e.g. that they can
all be reduced to some common kind of standard measure,
or that their variances can be stated in commensurable terms.
If, further, we propose to investigate objective differences
in variance, then we must contrive some method of equating
the units employed in measuring the variables to be com-
pared. Such problems have proved somewhat elusive in
actual practice, but they involve no new principles in theory.

In my view the most appropriate unit is not the standard devia-
tion, but the just-perceptible difference.1 In dealing, for example,
with English composition, after defining the group of observers and
the procedure to be followed, it is by no means difficult, though a
little laborious, to select a scale of specimens proceeding by steps
that are * just perceptible ' ; and further experiment will usually
show that, in such a scale, any two c equal-appearing intervals J
contain approximately the same number of * just-perceptible'

moreover are measurable " (S. J. Chapman, Elements of Political Economy,
pp, 60-1; cf. the mathematical treatment ofc utility * from Jevons and Edge-
worth to Bernadelli and Frisch). I may add that, if we take Meinong's dictum
literally, even * absolute value' may be regarded as quantity or magnitude.
" That is or has magnitude which allows the interpolation of terms between
itself and its contradictory opposite " (kc* cit.9 p. 8). Now, pleasure and
its contradictory unpleasure, ' positive value ' and its opposite * negative
value/ form a bipolar class; and in principle, provided we ignore the
qualitative differences between ' poetry and pushpin,' the bipolar class can
be converted into an ordered class in which a scries of terms can be inter-
polated between the two antithetical poles.

1 I believe that the success of standard measure in actual practice arises
largely from the fact that it so often converts crude measures into multiples
of the just-perceptible difference or into multiples of the causal difference.
As the Weber-Fechner law has taught us, where human judgment is con-
cerned, what we take to be equal additions are additions proportional to the
amounts to which they are added. In judging physical types, a tenth of an
inch added to the length of the nose makes as great an impression as a couple
of inches added to stature* Now, dividing by the standard deviation, like