ANALYSIS OF ILLUSTRATIVE GROUP 409

that suit our purpose best ? " l In the absence of a full insight into
causes, the best classification for purposes of diagnosis and prog-
nosis—though not perhaps of treatment—is a classification with a
minimum number of divisions and a maximum number of deducible
corollaries, each predictable with maximum probability. This, I
maintain, is given by the statistical procedure here pursued ; and
the same procedure will alone enable us to give definite answers to
still more pressing problems—how closely, for example, does the
behaviour of this or that person remain true to the predicted type,
in different situations, and in different moods, and at different
stages of his life ?

1 Printi-ples of Psychology, Vol. II (1901), p. 485.