290 THE FACTORS OF THE MIND

is adopted, then, when duly normalized, the factor loadings
for persons obtained by covariating persons are identical with
the factor measurements for persons obtained by covariating
tests. Similarly, the factor loadings for tests obtained by
covariating tests are, when duly normalized, identical with
the factor-measurements for tests obtained by covariating
persons ([101], p. 82 ; for proof and worked examples,
see Appendix II, Tables I-III). If the conditions specified
do not hold, the relations between the secondary factors
become a little more indirect, and vary according to the
nature of the changes : usually, however, the correlation
between the two sets of results is over -9 ([114], p. 166,
footnote r). Accordingly, it still seems true to say that, in
general, " except for minor differences of weighting, the
non-general factors obtained by correlating persons are
the same as those obtained by correlating tests or traits "
([137], p. 90). This conclusion is sometimes briefly
referred to as the * reciprocity principle.'

Here again my arguments have been questioned by Dr. Rhodes
and Dr. Stephenson, and once more from opposite standpoints.
Rhodes doubts the entire validity of applying the correlation calculus
to persons as variables ; but the results that he eventually reaches are
almost identical with those obtained by my least-squares method of
analysis from the corresponding correlations : hence, there can be
little real difference between us (see his further Memorandum for
the Examinations Inquiry Committee ([93], pp. 316-24)). Stephen-
son at first also expressed doubts about the method, but for different
reasons: it seemed to imply correlating measurements in entirely
disparate units, and would in any case produce (so he considered)
innumerable common factors—a result quite incompatible with the
two-factor theory to which he adhered (cf. above, pp. 177-8).
But, on reading the draft of my Memorandum dealing with Rhodes5
objections, he accepted the proposed mode of standardization and
the adaptation of Spearman's theorems, and also provisionally
agreed that the one method was " merely the obverse " of the
other—i.e. (as he puts it) the two procedures are merely " comple-
mentary ways of analysing the same data " ([92], p. 304; [96],

P* 195)-

As a test of these and other issues (more particularly the recipro-
city principle for abilities and temperamental tendencies) a
programme of researches was drawn up, and carried out in our