178 THE FACTORS OF THE MIND

been brought together in a monograph on Type-analysis,
to* be published shortly ; and I have to acknowledge his
kindness in communicating them to me in a series of letters
and memoranda. Like Dr. Rhodes, he was at first disposed
to reject the method outright. The difficulties he origin-
ally felt are best summed up in an early note criticizing
Dr. Dewar's first account of her investigation and com-
menting on a projected scheme for students' researches on
similar lines in the laboratory to which we were then both
attached. His objections are worth stating explicitly, not
only because they give the reasons for the important modifi-
cations that he was subsequently led to suggest, but also
because they voice the doubts so often raised by those who
are new to the proposal.

In a number of experiments carried out with the help of Miss
Bulley and Miss Felling, I had endeavoured to show, by correlating
the * marks ' given by different persons to sets of pictures, vases,
colours, etc., that " there was one general factor influencing the
artistic judgments of all," and in addition several less obvious
factors, producing more specialized types of appreciation (somewhat
similar to Bullough's types, and apparently related to more general
temperamental tendencies [75], 1933, p. 292). These conclusions
seemed confirmed, not only by minor investigations by earlier
students, but also by a long research carried out by Dewar with
school children, art experts, and unselected adults.

Stephenson, however, held that the statistical procedure adopted
in all these inquiries (a method of multiple factorization applied to
the correlations between persons judging) was " at once misleading
and futile " and " in conflict with the principles established by
Spearman." To start by correlating judges, instead of items tested
or judged, was, he argued, " misleading," because it suggested that
more factors entered into the tests than we know to be the case.
" With a hundred persons to be correlated, the number of inter-
correlations will be enormous ; and we shall consequently produce
a spurious increase in the number of factors " : when we correlate
by tests " we know from Spearman's studies that there will be only
one general factor in each battery," and no more than five or six
" over the whole range of the mind." Moreover, the procedure
" treats test-measurements, obtained in disparate units—a mere
heap of irregular and unrelated items—as a normally distributed
statistical population." Secondly, he considered the method to be