NATURE OF THE FACTORIAL TECHNIQUE 77

In all such cases, as may be seen from the table above,
each factor is characterized by what may be called a vec-
torial or columnar pattern. When the weightings are
fractional, each is distinguished by a unique series of
fractional figures for a certain relevant set of traits, that is
to say, by the saturation coefficients for those traits. And
each factor differs from every other, not merely by the set
of traits to which it is positively related, but also by the
relations between the trait-weightings themselves.

This is no new conception. Nearly all biological
classification rests on the same principle. We put all birds
in a class together, separate from mammals, fishes, and
amphibians, not in virtue of some single and simple avian
c essence '—their capacity for flight, for example—but
because their limbs and organs are differently related as
regards shape, size, and position from those of other verte-
brates—in a word, in virtue of their special mode of
organization—their characteristic structure, viewed as a
whole but with special emphasis on crucial points. The
relations beween their limbs and organs thus constitute a
general diagnostic pattern that defines the type.

The notion of a complex and distinctive attribute that
is expressed by a number of separate characteristics, and yet
is to be thought of as a unitary whole, is one that should
present no difficulty to the modern psychologist. The
word Gestalt expresses just this conception. When the
attribute is a mental factor, specified in quantitative terms,
the nature of the synthesis can be exhibited in graphic form.
If each contributory trait is represented by a vertical line,

deduced was checked by direct assessments made by schoolmasters. The
master's assessments for the different factors were by no means wholly un~
correlated with each other. Moreover, in the case, of the temperamental
factors the agreement between the master's assessments and the classifications
deduced from the tests was low. The results, however, were sufficient to
show that, from a practical standpoint, the construction of tests, which, like
the Binet scale, enable rough judgments to be made for a number of different
factors at a single examination, would lead to much saving of time and
labour. Such a conclusion seems fully in keeping with Fisher's view that, in.
experimental work, * excessive stress is often laid on the supposed importance
of varying only one factor at a time" ([109], pp. 100 et seq.9 § 37, * The Single
Factor').