CRITICISMS OF RESULTS 421

type or to that, " how is it," as Stephenson himself inquires, " that
it has been neglected for so long ? " ([98], p. 366). The answer is,
in the first place, that it has not been neglected so completely as is
commonly supposed. Davies, Moore, and myself used it for
assessing imagery-types as long ago as 1912 ; for determining types
of character, temperament, aesthetic appreciation, and the like, it
has often been employed in practical work ; x and in a recent survey
of the literature, over forty investigations are cited in which the
method of correlating persons had been used [130]. Nevertheless,
it must be admitted, the few theoretical investigators who have
relied upon it seem, as a rule, to have done so with reluctance and
apologies. For this the chief reasons have already been stated.
But there is one consideration that is especially pertinent here.
When we correlate persons, we are apt to introduce an over-
shadowing general factor, eliminated in correlating traits, which
may be quite irrelevant to our main issue. In certain inquiries, as
we have seen, this general factor may of itself provide the centre of
interest; but in researches on type-psychology, such, as the present,
it is an unnecessary and obscuring factor, and, what is worse, even a
fallacious factor—a kind of halo effect.

This has not always been recognized. But the effect can easily
be seen if a large number of judges are invited to grade or mark the
same person or persons for a list of traits such as that used in my
tables here. Some years ago, in The Measurement of Human
Capacities * I suggested, as an ' exercise in practical diagnosis,' that
readers should grade four delinquents (whose portraits were repro-

1 It was included, for example, in the summary of test-methods prepared
for the Board of Education in 1924 (Burt, ap. Report on Psychological Tests of
Educable Capacity, pp. 58-9). In its graphic form the principle has been
very freely used for constructing, either mentally or explicitly, a characteristic
* profile J to represent the ideal type or * standard personality/ and noting
how closely each individual's * profile ' approximates to it.

2 Oliver and Boyd, 1927, Year by year I have tried the same experiment
with my classes, sometimes at the beginning, sometimes at the end of the
course. In addition, each student is asked to rank himself and at least one
other person well known to him and to the group. In using portraits, my
object has been not so much to prove or disprove the value of physiognomy
as a guide to character, as to bring out the uses and limitations of impression-
istic judgments, when systematized by a scheme such as. the one we have
been using here. When the persons assessed are the assessors themselves,
the general factor for persons is still further coloured by the relative desirability
of each trait, which, of course, is much the same for all: no one, for example,
cares to call himself * flighty' or * fanatical *; hence in Stephenson's fists
these qualities sink towards the lower end of the scale for both his antithetical