ii2 THE FACTORS OF THE MIND

the illustration is not so remote as might be thought.1
Having measured each person's height, and the length,
breadth, thickness, and circumference of his limbs, head,
trunk, etc., we can express the observable resemblances
between the physical shape of the different individuals by
correlating the figures by persons. The object of a factor-
analysis will then be to discover the most comprehensive
principles by which we may classify and sub-classify the
sample population thus surveyed. Eventually we shall
obtain factors of the four kinds enumerated above, each

entering into them), I shall later show that the factor is the whole pattern or
system of traits rather than itself a further trait, added to, or underlying,
the rest.

As for the distinction of a c property ' in the strict Aristotelian sense, it is
now usually held that " in regard to organic kinds, the problem of dis-
tinguishing between ' essence' (i.e. the defining attributes) and * property'
(i.e. attributes co-extensive with the kind but not included in the definition)
is insoluble " (Joseph, Introduction to Logic, p. 102). With this, it will be
seen, I partly agree : or rather I would say that any solution is a matter of
convenience or convention. As in geometry, so in psychology, we might
take the * definition ' to be those essential characteristics which would enable
us either to construct the thing defined (as in dealing with tests) or to select
a typical sample (as in dealing with persons). The * properties' would then
be those further characteristics which (i) from factorial research we empiric-
ally find to be highly correlated with the essential characteristics and which
(ii) from other lines of research we can demonstrate (or at any rate plausibly
suggest) are consequences necessarily resulting from them. The whole sub-
ject, it will be seen, is closely bound up with the practical problem : how are
we to construct our tests, or to select our persons and traits, so that the most
useful factors will emerge from the statistical analysis ? Unless our defini-
tions tell us what to do in order to observe or measure the quantities defined
they are scientifically useless. For that reason I should prefer to base all
definitions—of traits, of tests, of sample populations, and of factors—on the
operations required to obtain them. This principle will come to the fore
later on when, it will be seen, we proceed to define our factors in terms of
selective operators (see below, chap, ix, § iv).

1 I may perhaps add that it has always seemed to me essential that the
demonstration of physical types, whether anthropological or temperamental,
should be based on a factor-analysis of the kind described in the text. It is
curious that this has never hitherto been undertaken. I have given the
results of a preliminary study in a recent paper ([114], pp. 184 etseq,}. An
illustrative analysis, by all possible methods, is given in my laboratory
Notes on Factor Analysis (CIL Physical Measurements5). It is impossible
to publish the numerous tables in full. Hence they are obtainable
only in roneo'd form.