CHAPTER XIX

FINAL CONCLUSIONS

A. MEASUREMENT OF TYPES

THE foregoing considerations show that, if we propose
to use correlation or factor-analysis for a scientific deter-
mination of temperamental tendencies, there are several
preliminary problems to be decided at the outset. Whether
we carry out a factor-analysis in full, or are content to
calculate the required correlation more simply and directly,
we are, in effect, correlating (i) measurements for the type,
which we take as a permanent standard of comparison, with
(ii) measurements for the particular person observed which
have been obtained ad hoc. Suppose, then, we are con-
templating a survey of temperamental tendencies among
the general population : two practical questions must be
answered first of all: (i) What special set of measurements
are we to take for the key personality or type F (ii) How
are we to secure individual measurements so that they shall
be fairly comparable with those for the type ?

(i) For the permanent standard, figures obtained from
a single batch, even if far larger and more representative
than the group here used to illustrate the modified pro-
cedure, could hardly be accepted without much wider
confirmation. The methods, however, first described in
my paper of 1915, have been applied to a number of mis-
cellaneous groups; and to the data originally summarized
a good deal of further material has since been added.
From each of the larger groups so studied I have calculated
a set of factor-measurements for traits, mainly by the
abridged procedure here described,1 and have then averaged

1 With the earliest groups the figures were obtained by correlating traits,
applying the summation method, and determining the factor-measurements
for persons by the usual method of regression. The results have since been
Checked both by the least-squares method and by correlating representative