i io THE FACTORS OF THE MIND

impression) to follow the various factor theories, if they are
presented to him first with reference to the classification of
persons rather than to the * analysis' (or, as I should prefer
to say, the classification) of tests or traits. Moreover, with
this alternative approach, many of the foregoing difficulties
disappear, or at least are simplified, because the selection
of a representative set of persons seems a much simpler
matter than the selection of a representative set of tests.1

The relation between the two approaches may be a little
clearer if we compare the essential logical nature of each.
When we factorize the correlations between the traits
tested, we are in effect analysing the connotation or ' inten-
sion ? of the class selected for testing ; when we factorize
the correlations between the persons, we are in effect
analysing the denotation or ' extension ? of that class. But
in either case, when we use one and the same set of observed
assessments to classify both traits and persons, it seems
obvious that, if our classifications are to be consistent, the
same factors must be preserved, the same jundamenta divi-
sionis retained, whether we start by comparing and classify-
ing the traits or by comparing and classifying the persons.

In each case the result of the analysis will be simply to
substitute a better defined and therefore more economical
set of attributes for the more numerous, more detailed, and
more casual set of attributes, superficially observable and
more easily measured, with which we set out. And, when
we turn from traits to persons, and proceed to study the
resemblances between individuals and to classify those
individuals according to the more orderly and pregnant
attributes ultimately adopted, it is not difficult to see that,
once again, our predications will be of four different kinds.
Nor is it surprising to discover that the four kinds of factor
empirically recognized by psychologists turn out to corres-
pond quite closely with the four or five headings of the
traditional scheme of predicables recognized by scholastic
logicians—genus, species (or differentiae), froprium, and
accident.

1 The experienced statistician will see that the problem of selecting a fair
sample of tests now gives rise to fresh theoretical difficulties; but these are
not likely to puzzle the beginner.