184 THE FACTORS OF THE MIND

On the other hand, with a view to surmounting the
objections to which (as he believed) our earlier work was
open, he proposes c radical modifications' in both the
experimental and the statistical procedure that had pre-
viously been employed ; and it is on these modifications
that he lays the greatest stress, since ' neglect of them in
the past has vitiated nearly all work on correlating persons,
and obscured its new possibilities.3 The experimental
procedure he proposes to alter by substituting " homo-
geneous tests " for the heterogeneous test material, which
was used by Bulley, Dewar, and myself, and which after a
preliminary trial he felt bound to discard : e.g. instead of
" a sample of 50 picture postcards representing all degrees
of artistic merit from reproductions of masters to the
crudest birthday card," he substitutes " 50 postcards of
Japanese vases all of approximately equal merit." With
this altered method of selection, he urges, the tests will no
longer constitute c a heap of irregular and unrelated items/
but may legitimately be treated as a * normally distributed
population.3 The statistical procedure he would modify
by substituting Spearman3s two-factor theorems (in
c inverted 3 form) for what he terms my ' multiple-factor
equations3 : the two-factor theorems, he contends, are
not only free from the c artificiality of a multiple-factor
technique such as your own and Dewar's, but also fit the
psychological facts more accurately.3 l

1 I quote from a covering letter that he was good enough to send me with
the draft of his first note to Nature, CXXXVI, p. 297, where he draws attention
to the possibility of using these particular theorems for correlations between
persons as well as between tests. In regard to the * multiple-factor equations'
there seems to have been some confusion. Stephenson and others apparently
supposed that these were put forward as a novel technique suited only to
correlations between persons. As he wrote in the same letter: " You
deserve great credit for trying to develop a new technique; but to me it
seems essential to use two-factor theorems." On the contrary, as will be
seen in a moment, my whole practice has been based on the assumption that
no * new technique ' was required. Actually, in * correlating persons * both
here and in America, two-factor theorems had already been used, where a
single general factor alone was extracted, and (in a few researches) for secondary
group-factors, where the persons correlated fell into discontinuous groups.
In the draft memorandum to which Stephenson alludes, the multiple-factor
method was preferred because it lent itself best to theoretical work in matrix