250 THE FACTORS OF THE MIND

species (or (Liferentice)^ proprium, and accident; and might
be regarded as a quantitative adaptation for the case of
variables of a qualitative scheme originally developed for
the case of attributes. More particularly, the * bipolar
factors' resulting from general-factor analysis correspond to
classification by progressive dichotomy, while the * group-
factors ' resulting from group-factor analysis provide an
equivalent classification of the same phenomena by co-
ordinate classes.

From the four-factor theorem (as it may be termed) all
the familiar factor theories may be derived. The differ-
ences, however, between the several kinds of factor are not
absolute, as these theories commonly assume, but merely
relative ; so that what is a group (or bipolar) factor in one
table may emerge as a general factor in another or a specific
factor in a third.

The primary value of such factors must obviously consist
in their utility for purposes of systematized description.
Whether or not any factor actually extracted or computed
happens to have a psychological significance is a problem
that must depend, not on the method of factor-analysis
employed, but upon the proper and relevant selection of
traits and persons. Factors cannot be invoked for purposes
of inference or prediction without additional data and
additional assumptions not included in the table of data
factorized or in the conclusions immediately drawn from it.
Indeed, the inferential and predictive use of factors as such
is far more limited than is commonly supposed ; and its
validity must rest upon a due observance of the general
conditions of all inductive inference—in particular, on the
stability of the results from one set of observations to
another and on certain a priori postulates about the subject-
matter of the inference, postulates which are seldom
explicitly announced in factorial work. Thus, the signifi-
cance and the reliability of the conclusions deduced will
turn quite as much upon the design of the experiment by
which the data are secured as upon the technique of the
analysis by which the data are examined. By itself factor-
analysis can at most describe only the general structure of
the mind or of the population : functional problems require