198 THE FACTORS OF THE MIND

sums up a distribution or pattern within one and the same
person, and c unipolar 'l in the sense that its saturation
coefficients in the different persons are always positive.
In investigations with cognitive tests, the ' general factor
for tests' is the factor obtained by averaging all the tests
for each person ; and in the same way the ' general factor
for persons' is the factor obtained by averaging2 all the
persons for each test. Thus, if we are chiefly interested in
comparing the tests, we can call this ' general factor for
persons ? the relative ease or difficulty of the tests (it will
usually vary with their complexity and level) ; if we are
chiefly interested in comparing the abilities, we can call it
the relative strength of the abilities in the whole population
tested (repeating eight numbers is a harder test than re-
peating four numbers because the average person's ability
to do the latter is greater than his ability to do the former).
In investigations on emotional traits—e.g. affective prefer-
ences or conative tendencies, the c general factor for per-
sons ' will similarly be the relative strength of those prefer-
ences or tendencies in the general population. And in in-
vestigations of almost every kind, since the resemblances
studied are complex, the c general factor for persons' will
(to put it briefly) describe a characteristic pattern of
measurements that distinguishes the universe of individuals
for whom the factor isc general' or generic : e.g. in human

is a technical expert, and (ii) his appraisal can be checked against those of
others ([136], p. 236). Nevertheless, these features by themselves do not
convert an individual grading into an objective measurement. On the other
hand, I admit that the underlying general factor (if any) is * objective ' in
that it is, by hypothesis, independent of any individual judgment. Indeed,
one of the purposes of my research was to show that artistic taste depends
largely on an objective factor in the same sense that weight-discrimination
does, though not to the same degree ([75], pp. 289, 294).

1 If I am right in regarding the table of positive correlations that is
commonly found in correlating tests as really a single quadrant of a larger
bipolar table (see above, p. 163), then the distinction between general and
bipolar factors would also disappear. This, however, is too speculative a
point to bring forward here, and would only confuse the issue.

2 The notion that a set of factor-measurements may be obtained by
averaging (weighted or unweighted) is another point in my theory to which
Stephenson emphatically takes exception (e.g. [98], p. 357 and ref.). How-
ever, this is a point which I hope to demonstrate in the last part of this book.