CRITICISMS OF RESULTS 417

his first factor and type are characterized by " consistent flow
of energy," and are " pushing, tenacious, headstrong," " fluent
in speech," and almost entirely free from " inhibition" or
" obstruction " ; the second factor and type have " inhibition "
and " obstruction" as their most well-marked traits, and are
" shut in and restricted." Surely the factors thus specified are
almost exactly the same as those underlying my own earlier
distinction between the uninhibited (aggressive or sthenic) type and
the inhibited (obstructed or asthenic) type respectively. What is
more, they surely have a striking resemblance to the factors
previously discovered and described by Miss Studman and himself
by correlating tests, and identified by Miss Studman with rny own.
However much, therefore, we may differ about procedure and
details, we seem after all to be in almost complete agreement about
the two main types that so constantly recur, and about their
essential characteristics.1

1 Unfortunately, owing largely to practical difficulties beyond the investi-
gators' control, in none of the experiments so far carried out has it been
possible to observe all the conditions that I suggested as necessary for a fair
comparison. These were " (i) that the group of persons used in correlating
persons should be the same as those used in correlating traits; (ii) that the
set of traits should be the same ; (iii) that both persons and traits should be
selected from their respective universes on an assignable basis, so as to form
(e.g.) an approximately normal, homogeneous, or rectilinear sample of the
* population'; (iv) that the same observer should make, or at any rate
standardize, the assessments for every person in the group." In the experi-
ment with hospital patients the observer was the same, but the trait-
assessments could not be normally or evenly distributed, because the group
was composed exclusively of extreme cases; in the two experiments with,
students, the trait-assessments might no doubt have been fitted to the
normal scale employed in other inquiries, but unfortunately the observers
were different: here, therefore, Stephenson has followed the same procedure
as Dewar and assumed that the * traits9 are normally distributed within the
same person : (this hardly seems legitimate with a list largely made up of
antithetical pairs; Dewar deliberately selected her * traits?—i.e. test-pictures
—so that they might reasonably be regarded as constituting a normal
distribution). These shortcomings are, of course, almost inevitable in
preliminary studies: as my attempts can testify, it is always very difficult
to find any one observer who can assess every student in a large group with
the same accuracy and detail as the medical officer in a mental hospital can
assess all his patients.

In the last of the illustrative experiments described by Stephenson—
1 experiment No. 4 : complementary R and Q analysis' (carried out by-
one of his students)—the difficulty was overcome by using test-
measurements for both types of analysis. The original intention was to

27