202 THE FACTORS OF THE MIND

particularly the psychological interpretation of the factors
themselves, many new problems arise, some as yet unsolved ;
but that does not of itself commit us to a c completely new
methodology.5 My provisional conclusion therefore is that,
although there may be material differences between the
factors obtained by the different ' techniques,' formally the
newer applications involve no fresh logical principles, and
lead to factors in no way radically different from those
obtained by the older.

Nevertheless, as I have from the outset insisted, when we
turn from the older applications to the newer, fresh light is
unquestionably shed both on the logical nature of factor-
analysis and on the intrinsic nature of the factors them-
selves. Though that light may reveal no novel features
hitherto unsuspected, it throws many of the previous
conclusions into sharper relief. Two important inferences
that are thus very clearly confirmed deserve amplifying at
somewhat greater length.

Factors as Principles for Classifying Tests.—In the com-
moner applications of factor-analysis, depending on the
correlation of tests, the investigator, as we have seen, usually
identifies his factors with abilities in the minds of the persons
tested : the ' factor-saturations for tests' are taken to state
how far each test depends upon, or is saturated with, the
particular ability specified, i.e. how far the results of the
empirical test would resemble or correlate with the results
of a perfect test for that ability. The * factor-measure-
ments for persons/ which are then deduced, indicate the
results that we should expect such a perfect test to yield,
i.e. they provide quantitative gradings or classifications for
the persons in respect of the hypothetical ability : thus, if a
particular examinee has a large negative measurement for
the ' general factor 3 of intelligence, exceeding some border-
line figure, he will for practical purposes be classified as
* mentally defective' ; similarly, if he has a large positive
measurement for the c verbal factor/ he can be classified
as belonging to the * verbal type.'

Now, when we adopt the alternative approach and start
by correlating persons, we reach similar c factor-measure-
ments 5 for tests instead of for persons. By analogy we must