4i8 THE FACTORS OF THE MIND

Since the above paragraphs were written, Stephenson has
published a further article x of great interest, in which he discusses
somewhat more fully the factorial analysis of temperamental types
by means of correlations between persons. The main contention
of my original paper [30]—that the chief temperamental types, as
described in non-statistical literature, can be verified statistically
and interpreted in terms of factors revealed by correlational
analysis—he accepts in principle : but he insists (i) that the
correlations must be correlations between persons, not between
traits, and (2) that, if types are to be established, these correlations
must be analysed by Q-technique, and by Q-technique alone. To
show how his method differs from that described in my last article,
he takes a set of temperamental assessments for 34 persons and
analyses the data in two alternative ways—first, by the two factor-
theorems of Q-technique and secondly by a multiple-factor tech-
nique, which he takes to be equivalent to my own.2 The argument
that a type is a " pattern of tendencies " and, as such, is capable of
unidimensional measurement, he emphatically rejects. Moreover,
he claims that the temperamental types he discovers are c limited

correlate the rows and the columns of one and the same complete matrix :
but the labour of calculating correlations for 100 persons seemed prohibitive ;
so only a selection was actually used for ' Q-technique.' Even so, the
correlation tables were presumably too large to print, and, indeed, no
figures at all are cited; but the investigator's own conclusions evidently
imply (as Stephenson's quotations indicate) that in this experiment the
two sets of factors were virtually, if not precisely, the same ([97], p. 360).
In his thesis cited above—A Statistical Study of Physical and Mental Types
(University of London Library)—the problem is discussed more fully,
with an illustrative calculation from physical measurements: it is found that
" the same results are obtained either by analysing correlations between
persons or by analysing correlations between traits." For the first bipolar
factor " a correlation of -90 was obtained between the factor-measurements
derived by factorizing traits and the factor-loadings or saturations derived
by factorizing persons " (loc. cit., pp. 213, 221).

1 *A Methodological Consideration of Jung's Typology,' J. Mental
Science, LXXXV, 1938, pp. 185-205. A more adequate answer to his
criticisms will, I hope, shortly be published in the same journal.

2 Actually the procedure does not conform to mine ; but that is doubtless
because, in my published articles, the abridged description of my method
was not altogether clear. His own procedure is to begin by selecting
4 or 5 persons representing each of his main types. Naturally, therefore,
his correlations reveal factors corresponding to these types, and his frequency-
distributions show discontinuous groups. If, however, we are allowed to
select our representatives in advance, surely we could demonstrate any types
we liked I