26 THE FACTORS OF THE MIND

warranted in allotting a higher a priori probability to the simpler
theory when we are dealing, not with the comparatively simple
phenomena of astronomy, but with the far more complex phen-
omena of mental life ?

A living creature is not a simple homogeneous ball, and its
relevant environment is as complicated as that of a planet is simple.
To suppose that the laws governing the movements of a planet
whirling in almost empty space can be expressed in three concise
equations containing only one or two terms and only low powers is
plausible enough ; but to expect that the laws governing the
movements of a wasp as it buzzes round a room will be equally few
and simple is the reverse of plausible. Again and again, the history
of pseudo-scientific theories in psychology has shown that the
commonest reason for accepting an erroneous explanation in the
past has been the strong popular prejudice in favour of simple and
single explanations where highly elaborate explanations would be
far more appropriate. The very simplicity of the * two-factor *
hypothesis has given it a widespread popularity among students
and teachers, but at the same time has led—or ought to have led—-
the neurologist, the biochemist, and the geneticist to be highly
sceptical of its finality. And much the same is true of most of the
speculative simplifications introduced by this school of psychology
or that.

This point of view gains some empirical confirmation from the
results of practical work. The three-factor theory of intellectual
abilities, which admits group-factors as well as the general and the
specific, is on the surface less simple than the two-factor theory,
which virtually excludes group-factors : yet in giving prognoses
for the development of subnormal children at a clinic, the former
appears to lead to far fewer errors, so far as can be judged from the
after-histories of the individual cases. We are told that the factors
deduced by the method of * principal components * are less in keep-
ing with the * law of parsimony J than the set of factors conforming
with the requirements of a £ simple structure '; and presumably
the same objection would lie against the method of least squares
(of which the method of principal components is a special form).
Yet, on re-analysing old data by various alternative procedures,
I find that predictions based on the method of least squares are
nearly always the more accurate.1 And, regarded as a basis for

1 This result was obtained by J. F. Steele and Miss E, R. Woodhead in a
recent unpublished research, where various modes of factor-analysis were
tried with the same set of clinical data, and the inferences checked by clinical
after-histories. As we apply it, the * method of least squares' or * weighted