P-, Q-, AND R-TECHNIQUES 177

so long as every person is assessed by the same observer, the
best and quickest method is still to proceed by correlating
traits, and to calculate factor-measurements for the several
persons (so-called c T-technique ?) l; if, however, each
person is assessed by a different observer, and particularly
if the persons are fewer than the traits or tests, then the
direct method is also the most practical, namely, to proceed
by correlating persons (so-called c P-technique ').1

2. The Unnecessary Multiplication of Factors.—The view
I have here put forward has been criticized still more strongly
by my colleague, Dr. Stephenson. His chief criticisms have

sociability, anger, curiosity, joy, and sex (roughly in that order, the reverse
of the order in which they appear in the typical introvert). We can thus
correlate the ranking for each individual with the ranking for a l standard
personality 9 as I called it—i.e. with a hypothetical set of marks representing
(in this case) the ideal extravert type : a positive correlation will then indicate
a tendency towards extraversion, a negative correlation a tendency towards
introversion, and the numerical size of the correlation the closeness of the
resemblance.

Copies of a graded schedule of traits have been regularly used for such
ratings and type-correlations, not only in class exercises with my students
(with the portraits on the lantern-screen), but also to obtain temperamental
estimates both of my clinical cases from teachers and of my own students
from each other. The * reliability * of such rankings is unexpectedly high
(cf. Burt and Spielman [5 3], p. 66). For exact research the method may seem
rather crude; but for the working purposes of educational and vocational
guidance it has proved eminently serviceable (see below, p. 426 f.).

If, instead of merely ranking the traits in order for each individual, the
reader endeavours to assign marks, he can correlate the marks by traits as well
as by persons. He will probably agree that this is a less reliable proceeding ;
but, even so, I fancy he will have little difficulty in convincing himself that the
correlation of each person with the * type * is roughly proportional (a) to the
factor-saturation obtained by factorizing the correlations between persons,
and (£) to the factor-measurement obtained by factorizing the correlations
between traits. However, these are points which I shall take up in Part III.

1 I plead guilty to using such shorthand phrases as P-axes, P-factors, and
even the P-method; but I do not care for the label It-technique, because, in
my view, there is no essential technical difference between, the analysis by
persons and by tests. P-factors and T-factors are convenient and colloquial
laboratory abridgments for * factors obtained by correlating persons* and
' factors obtained by correlating tests.9 By R-technique is to be understood
a special form of so-called T-technique, namely, * that which analyses
correlations between tests (or traits) by the application of the Spearman two-
factor theorems' (Stephenson).

12