3i4 THE FACTORS OF THE MIND

any rate become masked. In fact, except for one small
point, the new factorial matrix is exactly similar to those
obtained by some of us with what was termed the * group-
factor method 3 : it shows a step-like set of saturations
indicating three sharply distinguishable group-factors.
There is, however, a strange overlapping between two of
them, namely, between Thurstone's numerical and verbal
factors. His explanation is that the second set of verbal
tests (what I have called the miscellaneous sub-group)
show " some of the precision and restrictiveness of numerical
work " ; and he goes on to suggest that " the number
factor," which he supposes to be common to these three
verbal tests as well as to the four numerical tests, may after
all not be concerned with number as such, but with " some
kind of facility for logical or other restrictive thinking of
which numerical work is only a good example." That,
however, comes very near to expanding the numerical
factor into a general factor of relational thinking. This
third factor, indeed, has positive coefficients for nearly all
the tests: so that the attempt to dispense with a relatively
general factor is, after all, by no means complete or con-
vincing.

I venture to suggest that we have here an instance of the way in
which the assumptions prevalent among most factorists tend to
warp their psychological interpretations. These assumptions are,
as we have already seen, that a single factor must represent a single
ability and that it must therefore be interpreted in the light of its
positive saturations only. Such a view was natural enough in the
days when, owing to high probable errors, but one or two factors at
most could be extracted : for the first factor nearly always showed
nothing but positive saturations and could at once be explained as a
general ability ; and the second factor, which, as I should maintain,
really denotes a contrast between two abilities, could be identified
with whichever of the two had positive signs. This mode of
interpretation, however, is now extended to the third and fourth
factors that can often be distinguished with the more reliable data
of the present day. But in these further factorial columns we are
dealing with factors which draw a contrast, not between one pair of
abilities (or, as I should prefer to say, one pair of test-groups) but
between two or more pairs. If, following Thurstone's principles, we
eliminate, reduce, or otherwise ignore the negative saturations, then