378 THE FACTORS OF THE MIND

mental and aesthetic types in co-operation with several research-
workers at the Institute of Industrial Psychology and at the London
Day Training College: our conclusions were that the first or * general'
factor for persons (i.e. roughly speaking, the factor that is responsible
for the differences between those persons' averages) could not be
reached by correlating tests; but that each factor except the first (i.e.
the ( group-factors' or' type-factors') remained much the same for all
practical purposes, whether it was obtained by correlating persons
or by correlating tests (cf. above, p. 175). On resuming the work
at University College, I found my colleague, Dr. Stephenson, strongly
inclined to doubt—and, I think, quite rightly—whether these
inferences had been adequately proved. As a result of his own
experiments with the procedure, he was later led to suggest several
interesting modifications both in the methods to be used and in the
inferences to be drawn.1 His views in this connexion are deserving
of special attention, not only because they are based on formal
statistical analysis, but because he is one of the few statistical
psychologists with a wide experience of mental testing among
patients in mental hospitals. With his conclusions I shall deal later
on. Here I propose first to consider his criticisms in regard to
method ; then, accepting his suggestions as completely as I can, to
show how his own procedure leads in the end to virtually the same
results as mine.

His examination of my article is, I understand, appearing in a
forthcoming number of the same Journal; but he has been generous
enough to send me in advance a copy of his comments.2 Similar
questions have been put by other writers—psychologists, psychia-
trists, and statisticians ; and, if I deal more fully with the diffi-
culties advanced by Stephenson, it is mainly because, thanks to the
frequent opportunities for discussion, it has been possible to narrow
the points of difference or misunderstanding down to fairly definite
issues. Most of them, though not perhaps all, arise from the
incompleteness with which, for the sake of brevity, I was forced to
* describe the statistical procedure employed and the arguments on
which that procedure was based.

The chief objections may be summarized under four heads.

(l) Stephenson's strongest criticisms are directed against my use
of covariances instead of correlations. " Professor Burt," he writes,

1 The modifications have been briefly described above, pp. 182-8.

2 A brief statement of his main criticisms is to be found in the postscript
added to his article in Psycbometrika (III, pp. 206-9), which deals with the
more general proof given in [101]; a summary of the more specific points is
contained in his section of our joint article [138] : cf. also [136].