Personality

Personality
The psychometric view of personality is well established but little dealt

with in most textbooks. Personality: the Psychometric View sets out psy-
chometric methods and clearly describes the technicalities of testing and
factor analysis. In it, Paul Kline discusses different types of personality tests
and examines the main fmdings from the application of these tests and
methods. He further shows their utility in the applied fields of clinical,
occupational and educational psychology, as well as drawing out many
theoretical implications.
Personality: the Psychometric View is the only text on the psychometric
analysis of personality which is written for students, yet also deals with the
technical problems in this area, and reviews the work of all the main
researchers such as Cattell, Eysenck, Guilford and jackson.
Personality: the Psychometric View is designed for all students of psycho-
logy, education and the social sciences, as well as those in the medical
sciences who need to know about personality. Like its companion volume
Intelligence: the Psychometric View, it will also be useful to more advanced
postgraduates who need to use personality tests in their work.
Paul Kline is Professor ofPsychometrics at the University of Exeter.
Copyrighted Material
Other books by Paul Kline available from Routledge:
Intelligence: the Psychometric View

The Handbook of Psychological Testing
Psychology Exposed: Or the Emperor's New Clothes
Psychology and Freudian Theory: An Introduction
Fact and Fantasy in Freudian Theory (second edition)
Personality
The psychometric view
Paul Kline
~~ ~~o~:~~n~~~up
LONDON AND NEW YORK
First published I 993
by Routledge
27 Church Road, Hove, East Sussex BN3 2FA
Simultaneously published in the USA and Canada

by Routledge
270 Madison Avenue, New York NY I 0016
Reprinted 1994, 1997, 2004 and 2008
Routledge is an imprint of th e Taylor & Francis Group, an Informa business
© 1993 Paul Kline
Typeset in Bembo by Michael Mepham, Frome, Somerset

Printed and bound in Great Britain by
Biddies Ltd, King's Lynn, Norfolk
All rights reserved. No part of this book may be reprinted or reproduced or

utilized in any form or by any electronic, mechanical, or other means,
now known or hereafter invented, including photocopying and recording,
or in any information storage or retrieval system, without pennission in
writing from the publishers.
British Library Cataloguing in Publication Data

A catalogue record for this book is available from the British Library
Library of Congress Cataloging-in-Publication Data

A catalogue record for this book is available from the Library of Congress
ISBN 978-0-4 I 5-08978-4 (pbk)
Contents
1 The meaning of penonality 1

2 Measurement of personality: penonality tests 9
3 Factor analysis in the study of personality 32
4 The factor analysis of temperament 49
5 Findings from other types of psychometric tests 67
6 Penonality dynamics: the psychometric view 81
7 Heritability of personality 97
8 Personality testing in applied psychology 107
9 Penonality theory and the psychometrics of 131
penonality
10 Summary and conclusions 148
References 151
Name index 161
Subject index 165
Chapter 1
The meaning of personality
The term personality has many meanings in psychology. Indeed it has been
claimed by Hall and Lindzey (1957) that there are as many defmitions as
there are theorists. This is a serious matter for the scientific and systematic
study of personality since, clearly, definitions of terms affect the content
and method of what is studied. For example, the subject matter of papers
in psychoanalysis, especially modem research, bears almost no relationship
to that of situationalists such as Mischel (1968, 1977) or social learning
theorists, typified by Bandura (Bandura and Walters, 1963). In this first
chapter, therefore, it will be necessary to see to what extent there is
common ground within these different approaches to personality and to
delineate the trait account which underlies the psychometric view.
Different theories of penonality

It is not possible, or necessary, to describe in detail all personality theories.
Here I shall set out the main points of some of the most influential theories
and show how these theories have within them different definitions and
concepts. All this means, of course, is that unless considerable common
ground can be found, any research devoted to one theory will seem useless
from the viewpoint of another. That is why it is essential to clarify
definitions and concepts.
Psychoanalytic theories
Classical psychoanalysis, as typified by Freud (1939),Jungian theory Uung,
1940), more modem American psychoanalysis (Fromm, 1965) or the
recent French version (Lacan, 1966), has the unconscious, although
differently described, as a key concept. Psychoanalytic studies ofpersonality
must, difficult as this is, take this into account. This profoundly affects
2 The meaning of personality
method. Freudians stress the importance of defence mechanisms, such as

repression and reaction formation, the Oedipus and castration complex,
and the drives of sexuality and aggression. Child development is an
important aspect of the theory, especially psychosexual development, and
underlying everything is a closed energy model, instantiated in a mind
conceptualised as id, ego and superego, forces in fine equilibrium.
Jung's theories also embrace the closed energy model although the
descriptions of the mind are different and the aim of therapy and life
(analytic psychology is nothing if not bold) is not ego control but
individuation, a blend of penona and the wisdom of the collective
unconscious, attractive but certainly not attained by Jung and probably
unattainable.
ForJung and Freud, personality is seen as resulting from the equilibrium
of the mind which is conceptualised even in two such essentially similar
theories in concepts which are markedly different. For example, to study
the collective unconscious would throw little light on the Oedipus
complex. Clearly these concepts and definitions of personality demand
quite separate research.
Perhaps even more important is the fact that psychoanalytic theories in
general demand a way oflooking at the problems of personality which to
their adherents seem so natural as to be unnoticed. Thus the mind is seen
in terms of depth (the theories are sometimes called depth psychologies).
Generally the more deeply seated phenomena are the more important;
everything is caused; childhood experiences and fantasies are crucially
influential, and self-knowledge, on account of the unconscious, is im-
possible without analysis, except for the master. Such a psychoanalytic view
of penonality profoundly affects what appean to be important in the field
and thus the research and scholarship which might be undertaken.
This view of personality is in stark contrast to that of Bandura and
Walten (1963) whose social learning theory is essentially, as the name
suggests, a special application to personality of operant learning theory, as
propounded by Skinner (e.g. Skinner, 1953). For these behaviour is the
important component of personality. Mental events because they are not
public are not worthy of scientific study and, of coune, they are not
amenable to it. Phobias; for example, are regarded in this theory as learned
maladaptive responses. Personality is a set of learned behavioun. This is
different indeed from the psychoanalytic conception, of whatever school,
of a phobia as arising from deeply buried unconscious conflicts.
The research undertaken by social learning theorists into personality
will be concerned with establishing the patterns and conditions of re-
inforcement which are salient for penonality development and change.
The meaning of personality 3
These will be entirely different from psychoanalytically based research.

What is defined as personality by these two approaches has little in
common, although both see parents as important influences on personality.
One aspect of social learning theory which separates it entirely from
earlier psychoanalytic ideas must be mentioned briefly here although it will
be dealt with in detail later in this chapter. This concerns the scientific
method. Most modem personality theories attempt to be scientific in the
Popperian sense (Popper, 1959). This means that they are set out such that
they may be falsified. This demand, as shall be seen, has undoubtedly
affected the development of these theories and this is particularly true of
the psychometric approach.
Mischel (1968) originated what has been called situationalism, because
he argues that personality traits, far from accounting for personality, are
variables of relatively little importance because behaviour is far more
determined by the situation in which individuals find themselves. For this
reason, it is asserted, the correlations between personality traits and external
criteria are usually small. For example, a person may well appear
extraverted at a party or football match but will appear introverted, if
observed at a funeral or in a large library.
I shall not say more here about situationalism because its influence has
affected the modem psychometric account of personality, at least in
principle. However, there is no doubt that research driven by situationalism
is very different from that emerging from trait psychology.
Situationalism, because it attacked trait psychology and the psycho-
metric approach, is a useful entrance to the subject matter of this book, the
psychometric view of personality. However, before I tum to this it is
instructive to mention briefly yet another attempt to deny the importance
of personality traits. This is attribution theory which has been well
summarised by Eiser (1980). Essentially this asserts that traits are in the eyes
of the beholder. Traits are attributed to individuals as explanations of their
behaviour. They result from the cognitive processes of observers which
have to be studied rather than the actual behaviour of those who are
observed. The weakness of this position is that it denies the possibility that
traits are influential in determining behaviour. Logically both positions
could be true. Nevertheless, it illustrates the point, perhaps with even
greater clarity than could be done with the other theories, that definitions
of personality are intertwined with theoretical viewpoints.
Conclusions from descriptions ofpersonality theories

I have described the essentials of a number of personality theories to
demonstrate that the definitions and concepts of personality are dependent

on particular theories and viewpoints. Furthermore, theories ofpersonality
affect the kind of research and the methods of investigation which are used
so that there may be little in common between the results of different
approaches to personality. It would be surprising, therefore, if any one
theory were all embracing, although psychoanalytic theories do make such
claims. Clearly, however, given these problems it would be advantageous
if a particular approach to the study of personality could cover much of
the ground.
Trait theory of penonality

As Kreider and Kreider (1990) point out, the everyday conception of
personality is in terms oftraits. Traits are conceptualised as stable tendencies
or characteristics of individuals. Since it is obvious that people differ in
terms of traits, it is natural that the psychometric view of personality should
be a trait theory, for psychometrics is the study of individual differences.
The psychometric view of personality, therefore, arises from the study of
individual differences in personality traits.
Personality traits are used, in trait theories, to answer two fundamental
questions, those concerning the determinants of behaviour and the struc-
ture of personality - how traits are related. Thus to a trait theorist
personality is the sum of an individual's traits and these traits explain that
person's behaviour. The research questions, therefore, in trait psychology
involve the number and nature of personality traits and their relations to
behaviour.
The psychometric view of perSonality, or what might be called the
psychometric model of personality, constitutes the answers to these two
questions which have been obtained from the psychometric study of
personality. It is these answers and the methods from which they were
derived which form the core of the present book.
It should not be thought that the psychometric view of personality and
the trait view are identical. This is not the case because it is possible to
develop a trait view without psychometrics, or only partly based on
psychometrics, as was done, for example, by Murray (1938). McDougall
(1932), indeed, expounded an important and influential trait theory
without tests of any kind. However, as shall be argued throughout this
book, the psychometric view of personality can offer a greatly improved
version of the trait model, because it is based upon sound measurement.
The psychometric model of personality
Definition ofpersonality
In the psychometric model, personality is defined as the sum of an
individual's traits which determine all behaviour. Thus, as shall be seen
later in this book, it is possible to write specification equations for a variety
ofbehaviours in terms of traits. In some cases good predictions can be made
from them. Before explicating this model in the remainder of this book,
it is necessary to point out some of its advantages compared with other
models or views of personality.
The psychometric model and scientific method

Throughout this book it is assumed that the best method to obtain valuable
knowledge about human personality is through the scientific method. This
is not an article of faith but is derived from the fact that in the natural
sciences the application of the scientific method has led to an enormous
increase in understanding. However, the subject matter of psychology
differs so considerably from that of the natural sciences that it is arguable
that the scientific method is not well suited to it. Indeed I have demon-
strated (Kline, 1988) that in many branches of psychology the scientific
method, at least as conceived by its practitioners, manifestly fails. However,
in the field of personality, the scientific method, as represented by psy-
chometrics, seems able to make some progress.
Essence of the scientific method

As was discussed earlier in this chapter, a critical aspect of the method lies
in the formulation of refutable hypotheses. In practice this has demanded
that research in the field of psychology has certain characteristics which are
set out below.
-All variables should be quantified. Of course quantification can take
various forms and the higher the precision the better.
-Samples should be large and representative. On account of the
heterogeneity of human beings and their enormous numbers samp-
ling is essential in all experimental work. Clearly samples must be
adequate if conclusions beyond the experimental results are to be
drawn.
-Statistical analyses should be carried out. It is essential that statistical
analyses be carried out to show to what extent the results could have
arisen by sampling error. This is the problem which arises from most
clinical studies. However, as will be fully discussed in the relevant
sections ofthis book, many statistical analyses reported in the research
literature are so poor as to be misleading. The reverse of this error
is also often found, as has been shown by Kline (1988). Here
researchers are so determined to produce sound statistical work that
they choose problems easy to analyse but of a profound triviality.
-Research designs should be such as to allow proper conclusions to
be drawn. This aspect of the scientific method is closely related to
the statistical analysis discussed above. However, in many studies of
psychotherapy, for example, no control groups who receive a
placebo treatment are used. This makes evaluation impossible.
-Hypotheses should be drawn up such that they may be refuted. This
is the critical aspect of the scientific method. However, only by
ensuring that the four points above are properly executed is it truly
possible to refute hypotheses.
Nevertheless, despite its apparent simplicity, a few comments should be
made about the principle of refutability. First it means that no scientific
knowledge is fixed but is always held true until it is refuted. Furthermore,
it should be pointed out that hypotheses cannot be proven, only refuted.
An obvious example of this can be seen with the hypothesis that all swans
are white. This hypothesis can never be proven no matter how many white
swans are observed. It may be held until refuted by observing a black
specimen. It should further be noted that a hypothesis need only be
logically refutable to be scientific. Thus before the development of space
rockets it was quite scientific to hypothesise that the moon consisted of
cheese or any other substance since this was, in principle, testable.
Finally there is another, perhaps more fundamental, difficulty associated
with the notion of refutability. This concerns its meaning. As Gruenbaum
(1984) has argued, careful analysis indicates that the only meaning which
can be attached to the claim that some hypothesis is not refutable is that
the individual making that claim cannot think how it might be refuted.
Nevertheless, despite this problem, as an effective, practical approach to
carrying out scientific research the Popperian notion ofscience as depend-
ing upon the formulation of refutable hypotheses is highly valuable.
From this analysis of the scientific model it can be argued that psy-
chometrics and, therefore, the psychometric model is well able to meet
the demands of good scientific work. Thus psychometrics is concerned
with the development of psychological tests thus ensuring as sound

quantification as possible. Similarly psychometrics has always placed great
emphasis on sampling, research design and statistical analysis, as can be seen,
for example, in Nunnally (1978), Cattell (1978) and Kline (1992a). This is
one important advantage of the psychometric model of personality com-
pared with other approaches. Thus it is notoriously difficult to put
psychoanalytic theory to a rigorous scientific test, although by no means
impossible (Kline, 1981).
There is a further advantage inherent in the psychometric view of
personality. This is simply that it is virtually all embracing. It claims that
all behaviour can be understood in terms of traits. If at present this seems
to be impossible, this is only because of shortcomings in measurement and
in the mathematical conceptualisation of the problems of personality, at
least as argued by Cattell (1981).
Before concluding this chapter a little more needs to be said about traits.
In the English language there is a huge number of trait terms. Traits which
seem to be concerned with problem solving, such as intelligence, a
pervasive trait, are usually conceptualised as ability traits. These have been
shown (Cattell, 1957) to be separate from personality traits in the sense that
the correlations between traits in these different fields are generally low
and non-significant. However, there is a distinction to be drawn among
personality traits themselves. This is between temperamental and dynamic
traits. The former account for how we do what we do, the latter for why
we do it. To exemplify the distinction, extraversion is a temperamental
trait. Extraverted individuals can be easily spotted. As they do things they
make a lot of noise and expend a great deal of energy. If other people are
about they talk and joke with them. Dynamic traits, on the other hand,
are drives. Fear can be a powerful drive as is sex or hunger.
A distinction can also be made between states or moods and traits. Traits
are relatively constant, enduring characteristics of an individual, whereas
states are transient. Anxiety is particularly interesting in this respect because
there is clearly trait and state anxiety. Trait anxiety is the general level of
anxiety which each individual has, if nothing particularly arousing has
occurred. State anxiety reflects the anxiety provoked by some event or
thought. Visits to the dentist, examinations and unusual situations are all
likely to arouse state anxiety.
In the relevant chapters of this book, these distinctions will be fully
discussed and the basis for them will be examined together with their
implications for the psychometric view of personality. Here it is sufficient
to note that they exist.
Conclusions
In this chapter I have described briefly a number of different theories or
views of personality to demonstrate what is inevitably the case that
definitions and concepts used in the study of personality depend upon the
theory involved. This is important because there are many different
theories of personality, with relatively little overlap. In this context the
psychometric model of personality was described and its relation to the
more general trait model was explicated.
It was shown further that the psychometric model of personality was
consonant with the scientific method as applied to the study of personality
with its demand on precise quantification and the clear formulation of
testable hypotheses. It was also shown that the psychometric model of
personality is comprehensive in its coverage since it claims to be able to
account for all behaviour in terms of traits.
The remainder of this book will be concerned with a description and
scrutiny of the psychometric view of personality, an examination of its
application in various applied fields of psychology and finally of its
contribution to psychological theory and knowledge. In Chapter 2 the
measurement of personality is discussed.
Chapter 2
Measurement of personality
Personality tests
Since the early days of psychology, there have been attempts to measure
personality, with a variety of different kinds of tests. As a result of this
considerable research effort, in the modern study of personality there are
now three kinds of psychological test in general use: personality question-
naires or inventories (these terms are absolutely interchangeable);
projective tests; and objective tests.
Before describing these different types of personality tests, a few more
general points about psychological measurement need to be made, thus
enabling a fine examination of personality measurement.
The characteristics of good psychological tests

Psychometrics is the study of individual differences by means of psycho-
logical tests. Psychometricians, as specialists in measurement, attempt to
produce tests with certain key features and these are set out below. They
are the essentials of precise measurement and are thus critical for the proper
application of the scientific method.
All good tests should be highly reliable, valid, discriminating and have
good norms. The meaning of these terms will now be discussed. To do
this, however, a number of concepts require some explanation.
Universe of items Any set of items in a test is assumed to be a random

sample from a universe of relevant items. This universe is, of course,
notional and infinite. The better the sample of items the better the test.
The score on a psychological test is known as the obtained or the fallible
score. This is to be distinguished from the true score. Any fallible score
consists of true score + error.
True score The true score consists of the score of a subject on the
10 Measurement of personality
univene of items. This is, therefore, a notional score. However, as shall be

seen, it can be estimated from the obtained or fallible score provided that
the reliability of the test is known. Since the square root of the reliability
of a test (its homogeneity and its dependability over time, but see below)
is an estimate of the correlation of the test with the true score, it can be
seen that the higher the reliability the less the error of the test and the more
closely the obtained scores approximate the true scores.
All this is part of classical test theory and is expounded with brilliant
clarity in Nunnally (1978).
Reliability
Reliability has two meanings which will be scrutinised separately. One
concerns the internal consistency of a test, the other its stability over time
-known as test-retest reliability.
Internal consistency reliability Any measure, not simply a psycho-

ogical test, should be internally consistent. If different parts of it are
measuring different variables it is hard to see how it could be a good test.
For this reason internal consistency is regarded as a desirable attribute of
tests. It should also be pointed out that in psychometric theory (see
Nunnally, 1978 or Kline, 1986) it can be shown that reliability is inversely
related to error of measurement. The only dissenting voice among leading
psychometrists, is that of Cattell (1973) who argues that high internal
consistency can lead to the measurement of rather n.arrow and psychologi-
cally trivial variables. His ideal is a test in which all items correlated with
the variable which they were supposed to measure but not at all with each
other. There is good logic in this argument but in practice it is almost
impossible to construct tests of this type. In the view of this writer, high
internal consistency reliability is essential but not too high because of the
dangers of measuring trivial variables, which undoubtedly occur in social
psychology, for example, as has been shown in Kline (1992a).
How reliability is measured must now be discussed. The simplest
method is to split the test in half and correlate the two sets of scores - the
split-half reliability. However, since a test can be split in many ways this
may be inaccurate and in fact Cronbach's alpha (Cronbach, 1951) is the
preferred coefficient. This provides an estimate of the correlation of the
set of test items with another set of similar items from the same universe
of items. The square root of alpha is the estimate of the correlation of the
test with the true score, as was pointed out above. This makes it clear why
internal consistency reliability is regarded as so important. The higher it is
Measurement of personality 11
the less error. In fact an alpha of0.7 is regarded as a minimum figure for
an adequate test.
It might be thought that there is some confusion, or even contradiction,
in the first claim that tests must be internally consistent since the higher
the reliability the less the error, and the argument of Cattell, that internal
consistency should not be too high. However, this is not the case. Tests
can be made highly consistent by using items which are virtual paraphrases
of each other. These are highly reliable, but in terms of classical test theory,
they are samples ofa universe ofitems oflitde psychological interest- items
which are semantically similar to each other.
Classical test theory is statistical not psychological, with the result that
it is concerned simply with universes of items but is uninterested in the
psychological meaning of these universes. If we consider a universe of
extraversion items, it becomes obvious that this is broad and that items
which are genuinely part of it may not correlate highly with each other.
Thus both Cattell and classical test theory are correct and in practice it
seems best to aim for tests with alphas beyond 0.7 but to be suspicious,
especially in the sphere of personality as distinct from ability, where alphas
can be high, of alphas beyond 0.9. This problem of reliable but narrow
and specific tests oflitde psychological interest will be dealt with in more
detail in the next chapter on factor analysis, and will occur again later in
this chapter under the heading of validity.
Test-retest reliability There is no need of abstruse psychometric theory

or recourse to notional true scores, to understand the psychometric demand
that the test-retest reliability of tests be high and again 0. 7 is a minimum
figure for a satisfactory personality test. It is perfecdy obvious that if a test
is given on two occasions and gives different scores each time that one, at
least, must be wrong and that no trust could be placed in either, provided
that the subject tested had not changed.
All this is so banal as, perhaps, to appear hardly worth writing. Yet two
of the most common forms of assessment, the interview (used in occupa-
tional selection) and the essay (used in education), have notoriously poor
test-retest reliability (Vernon, 1964), in some cases litde better than zero.
Standard error of measurement

As has been argued above, obtained scores from tests are estimates of true
scores. The higher the test-retest reliability the closer to the true score are
the obtained scores. The standard error of measurement estimates at
different levels of probability what the true score of an individual might

be.
Formula for the standard error of measurement, O'meas = O't /.J1-ru
where O't = the standard deviation of the test and rtt = test-retest
reliability.
The standard deviation of a test is the index of the variability of the scores.
In a normal bell-shaped distribution of scores 68 per cent of all scores fall
between the mean (average) and plus and minus one standard deviation.
95 per cent ofall scores fall between the mean and two standard deviations.
For example, if a test has a mean of 50 and a standard deviation of 10:
68 per cent of scores fall between 40 and 60 and 95 per cent between 30
and 70.
Importance of the standard error of measurement The standard

error of measurement is highly important if test scores are to be used as a
basis ofjudgement about individuals. This standard error is the estimated
standard deviation of scores of an individual taking a large number of
parallel sets of items. It can thus be used to set confidence limits around a
test score. An example will clarifY the point and demonstrate its value.
Suppose that we have a personality test score of a subject of 120 and
further suppose that the standard error of measurement is 5. Then 68 per
cent of that individual's obtained scores would fall between 115 and 125
and 95 per cent would fall between 130 and 110, if she were to be tested
again and again. These are the boundaries of the true score for that
individual. Given these standard errors it would be ridiculous to select her
for a post over another candidate who had scored 116.
Again if we were comparing the scores of one person on a variety of
subscales of a test, differences would have to be beyond the error boun-
daries to be taken seriously.
The implications of these points are obvious. For all practical testing it
is essential that the test be highly reliable. A test with poor test-retest
reliability is of little value because its scores give a poor indication of the
true score.
Factors influencing test reliability 1 am stressing the importance of

reliability because, as shall be seen, many personality tests, especially
projective tests, are so unreliable that they can be used only with the greatest
caution. Factors which are likely to make tests reliable include objective
scoring, where no judgement is required by the scorer and having a large
number of items. It can be shown that the more items a test has the more
reliable it is, until it becomes so long that boredom and fatigue set in. This
is the Spearman-Brown prophecy formula (see Nunnally, 1978 or Kline,
1992a). A twenty-item test can be highly reliable but a reliable test ofbelow
ten items is probably too specific to be useful.
One further point highly relevant to the study of personality needs to
be made. Some variables such as moods and states fluctuate quite consid-
erably. Thus the test-retest reliability of measures offear or anger is likely
to be low. While it is true that a measure of anger given on two occasions
might have a low correlation, this is not necessarily on account of low
reliability. This is because the status of individuals on the variable has
changed. A good test ought to register differences. Reliability is concerned
with changes due to error, not to function fluctuation, as it is called by
Cattell (1973). The correct way to estimate the test-retest reliability of a
variable such as anger would be to arouse the anger in subjects on each
occasion of measurement.
However, important though test reliability is, as has been shown, it is
so only because it contributes to the validity of tests and this must now be
discussed.
Test validity
A test is valid if it measures what it claims to measure. However, with this
definition all depends upon how what a test measures may be demon-
strated. This difficulty has led to a number of different types of validity and
these will now be described. It should be pointed out at this juncture that,
unlike reliability, there is no single figure which indicates test validity.
Indeed, according to some writers, e.g. Vernon (1964), a test is valid for
some particular purpose or with some particular group. Thus a test might
be valid in the selection of military personnel but not useful for doctors.
Face validity This refers to the appearance of a test. If it looks valid it

has face validity. Unfortunately, with personality tests, there is no necessary
connection between face validity and true validity. Indeed, the only
demand that tests have some degree of face validity is that without it
subjects may not cooperate in the testing. This unquestionably lowers the
validity of the test. However, if a test is truly face valid, it may lead to
distortion, especially in selection. Candidates for the armed services would
be unlikely to admit to physical fears, for example.
Concurrent validity To demonstrate concurrent validity, a test is

correlated with another test of the same variable, both tests being
administered at the same time. For satisfactory concurrent validity a

correlation ofat least 0.70 between the two tests would be expected. There
are several points to be noted about this type of validity. The fint concerns
the logic of the procedure. If there is a good benchmark test of the variable,
then a high correlation is convincing evidence of validity. However, in the
field of personality this is rarely the case and only the measures of
extraversion and anxiety in the EPQ (Eysenck and Eysenck, 1975) could
be regarded as satisfactory for this purpose. Nevertheless, even here there
is a problem. If these scales are so good what is the point of another scale?
Only if the new scale has real advantages, for example, it is easier to
administer, or much shorter than the benchmark scale, is there a point in
trying to develop it. Obviously if scales are not themselves highly valid, a
high correlation would not be convincing and a moderate correlation
would be difficult to interpret. Hence concurrent validity is not much used
on its own. However, as one piece of evidence it can be useful. This is
discussed below, under the heading of construct validity.
Predictive validity A test is said to possess predictive validity if it can

predict some relevant outcome. For example, if a group of people were
given an anxiety test, they could be followed up a few years later and the
anxiety scores could be used to predict psychiatric breakdown or treatment
for psychiatric or psychological problems. If there were a significant
correlation, the test would have predictive validity.
This would be an impressive demonstration of validity as predictive
validity always is. However, in the case of personality tests it is difficult to
set up a good criterion, with the exception of anxiety, so that, as with
concurrent validity, it is little used, except as an aspect ofconstruct validity.
Incremental and differential validity These two types of validity are

most frequently found in occupational psychology. Incremental validity
can be claimed when a test correlates rather low with the criterion score
(success at a job, for example) but zero with other tests in the test battery.
Such a test is valuable when the multiple correlation between the test
battery and the criterion is computed. This will be further discussed in
Chapter 8.
Differential validity occurs when a test correlates differentially with
different parts of the criterion score. The most usual example of this is to
be found in the prediction of academic achievement. Thus intelligence
correlates much the same with degree classes in all faculties while interest
tests and tests of extraversion have different correlations with different
subjects. Vernon (1950) contains a useful discussion of this point.
Construct validity Because, as has been seen, it is difficult to establish

the validity of penonality tests by any of the methods which have so far
been discussed, construct validity is usually the chosen approach. The
concept of construct validity was developed by Cronbach and Meehl
(1955). The demonstration of construct validity requires that hypotheses
concerning the nature of the test variable (construct) be set up. If all of
them, or the majority, are confirmed, then the test may be said to be valid.
An example, an anxiety test, will clarify the nature of construct validity.
The following hypotheses would be tested:
1 High scoren would be more likely to be receiving psychiatric treatment
than low scoren.
2 High scoren would be less likely than low scoren to be in stressful or
dangerous occupations.
3 The anxiety test will have positive correlations with other anxiety tests.
4 The anxiety test will correlate zero with penonality tests not measuring
anxiety.
5 The anxiety test should correlate zero with tests of ability and motiva-
tion.
If all these hypotheses were supported it would be difficult to argue that
the test was not measuring anxiety. From the nature of the construct of
anxiety, these are the results which would be expected. It is to be noted
that construct validity involves a pattern of results, a mosaic of findings,
and this is the cause of one of the difficulties with the concept. In practice
results are not so clear cut that all or none of the hypotheses are confmned.
Thus there is a strong subjective element in the assessment of construct
validity. Claims for construct validity must be carefully scrutinised.
Conclusions concerning validity It is clear from this discussion that

the validity of a test is a subjective issue which cannot be settled by the
production ofsome dear statistic. With the exception offace-validity, each
type of validity bean on the question of whether the test measures what it
claims to measure although differential and incremental validity are more
concerned with the practical utility of the test. Construct validity is the
summation of all the findings.
Validity has been discussed at some length because, as will be apparent
throughout this book, the psychometric approach to penonality seeks to
investigate penonality by means of valid tests while many of the other
penonality theorists use assessment methods and tests oflow validity (where
they even attempt to measure variables) which renden the work useless.
Discriminatory power
This is the third characteristic of good psychological tests which can be
dealt with briefly. One of the aims of a good test is to produce a spread of
scores - discriminatory power. This is obviously essential for psycho-
metrics, conceived of as the study of individual differences, and is equally
obviously important if we consider the value of a test on which all subjects
score the same. Given equal reliability and validity the most discriminating
test would be best.
Discriminatory power is measured by Ferguson's delta (Ferguson, 1949)
which runs from 1 (maximum discrimination) to 0. A normal distribution
has a delta ofabout 0.9 which is perfectly satisfactory. The maximum value
is attained by a rectangular distribution where the same numbers ofsubjects
are found at each score. In practice, of course, such a distribution is almost
impossible to obtain.
The reason for discussing discriminatory power is that personality
questionnaires and inventories usually have high deltas while other assess-
ment techniques such as the interview or rating scales are poor
discriminators. Vernon (1950), for example, showed that raters could hold
a maximum of nine categories in their heads while interviewers could do
little better than use three categories: average, below average and above
average.
Test norms
The fourth characteristic of good psychometric tests is the possession of
good norms. Norms are sets of scores from clearly defined samples and the
setting up of these norms constitutes test standardisation.
Without norms the psychological significance and meaning of an
individual's score on a test is unknown. This is, of course, because, unlike
many measures used in the natural sciences, personality tests have no
meaningful zero. Thus a score of 20 on an extraversion test, for example,
is interpretable if it appears from the norms that such a score is exceeded
by only 2 per cent of the population.
There is no need to discuss standardisation in any detail here (see Kline,
1992a for further information) . It is sufficient to note that the samples on
which the norms are based should be large and representative. If they are
not, norms can be misleading, worse indeed than useless. The actual form
which norms take varies considerably with different tests. The most
commonly used are T scores with means of 50 and standard deviations of
10.
Summary and conclusions

Four essential characteristics of good tests, high reliability, validity and
discriminatory power, together with efficient standardisation, have been
discussed. In the light of this discussion the different kinds of personality
tests will be described and scrutinised. However, before this is done one
distinction needs to be made. This concerns the difference between
nomothetic and idiographic measurement.
1 Nomothetic tests. These are concerned with variables common to
individuals. Tests of extraversion or obsessionality fall into this category.
2 Idiographic tests. These seek to measure aspects of personality unique
to individuals. An example of such a test would be one aimed to assess
the nature of a subject's unconscious conflicts, which, according to
psychoanalytic accounts, would be unique.
In principle any type ofpersonality test could be nomothetic or idiographic
although in practice personality questionnaires and objective tests fall into
the former category whereas projective tests tend to be idiographic.
TYPES OF PERSONAUTY TEST
Personality questionnaire or inventories

As the names suggest, these tests consist of sets of items, which are usually
questions or statements relating to thoughts, feelings or behaviour. Subjects
are required to respond appropriately to these items. The most common
types of item are illustrated below.
The Yes/No item Must you be in plenty of time if you have to catch
a train?
Yes/No.
This is a useful form of item, found in the EPQ (Eysenck and Eysenck,
1975), since it is relatively simple to write and is applicable to a wide range
of behaviour and feeling. However, this simplicity can be regarded as a
problem. Thus Heim (1975) regards such items not as simple but simplistic
(thus insulting the intelligence of subjects and creating a poor attitude to
taking such tests) and certainly it is difficult to catch the full subdety of
human behaviour with such items.
There is a variation on this item in which there is a third category for
subjects to use if they are uncertain. However, this category may be too
attractive for some subjects, although it is not highly informative, with the
result that the questionnaire will not be as accurate for them as it should
be. Since there is a substantial correlation (Bendig, 1959) between these
two fonns, however, it becomes simply a matter of preference which is
chosen.
The True/False item I regularly feel sick before exams.

True/False.
This is an item form (used in the MMPI, Hathaway and McKinley, 1951)
highly similar to that already discussed. These items usually consist of
statements in the first person to which subjects must respond.
The Like/Dislike item Baked beans.

Like/Dislike.
This type of item consists or words or phrases to which subjects have to

indicate like or dislike. Clearly all depends on the choice of words and this
itself reflects the theory underlying the variables measured. This is not a
common form of item but it is used in the DPI by Grygier (1975). This
test is derived from psychoanalytic theory and thus contains items of food
(oral fixation) and phallic symbols (phallic fixation), just for example.
Items with rating scales People should be more self-controlled.

Strongly agree IAgree I Uncertain I Disagree I Strongly disagree.
These items consist ofstatements to which rating scales, 5, 7 or 9 categories,
are appended. Depending on wording different scales can be used, for
example, always to never. In the field of personality Comrey (1970) is the
keenest proponent of this type of item because the rating scales allow more
precise correlations between items, as is discussed in Chapter 3. In addition,
these items are not so obviously simplistic. However,. there are two
difficulties with these scales. Some subjects tend to avoid the extremes
while others use them frequently. Furthermore, there are bound to be
differences among subjects in how they interpret the meanings of the scale
tenns. Finally it should be pointed out that some attitude tests are of this
type and are known as Likert scales (Likert, 1932).
Various trichotomous forms These are variants of the Yes/No items

which have been discussed. Sometimes, for the sake of good sense, the
following fonns may be used: generally, sometimes, never; true, uncertain,
false; agree, uncertain, disagree.
Forced-choice items These involve competing phrases of which sub-

jects have to choose one. An example item: When I'm feeling really tired
I like to: (a) relax in a hot bath; (b) watch the TV; (c) work out in the gym. The
number of choices can vary from 2 to 5 or even more. These items can be
irritating if subjects feel that they hate all possibilities.
Edwards (1959) was a powerful advocate for forced-choice items. He
matched the pairs for social desirability in an attempt to eliminate the
tendency to endorse items according to how socially desirable it was so to
do. This response set will be discussed later in this chapter.
Ipsative scores With forced-choice items in which each choice receives

a score on a different scale, resulting scale scores (known as ipsative scores)
are negatively intercorrelated. Furthermore, they relate to the relative rank
of each scale for each individual taking the test. This means that strict
comparison between individuals is not meaningful and norms should not
be used. Furthermore, because the scales are artefactually correlated,
correlational analyses are impossible to interpret. This rules these tests out
for factor analysis (see Chapter 3).
Of course not all forced-choice items are ipsative. If a subject scores 0,
1 or 2 on the same scale depending on her response, the items are entirely
suitable for further statistical analyses and comparison with other scores.
In brief, this means that tests with ipsative scores are only suited to
discussion with the individual who completed the test and are not useful
in the psychometric study of personality.
Although there are other types of item used in personality inventories,
the types discussed above embrace the vast proportion of items ever found
in tests. From this discussion, the nature of personality questionnaires is
clear. I shall now tum to a briefdiscussion of their advantages and problems.
Problems with personality questionnaires

There are certain problems with the use of personality questionnaires
which can be discussed briefly because, in general, careful test construction,
which will be examined later in this chapter, can overcome them.
Frequently mentioned as difficulties are response sets which determine
subjects' answers to items and thus lower the validity of the test. The main
response sets are discussed below.
Acquiescence This is the tendency to agree with responses, regardless

of their content. Vague items tend to produce this response. One method
of dealing with it is to have half the items keyed 'No', thus ensuring that
acquiescent individuals are not confused with high scoren on the test.
Actually, in test construction items which attract acquiescent responses can
be eliminated.
Social desirability This has already been mentioned in connection with

forced-choice items. It is the tendency discussed by Edwards (1959), to
endorse items on account of their social desirability. Thus he found a high
correlation between rate of endoning an item and judgements of social
desirability. That is why he used forced-choice items balanced for this
variable. Unfortunately, however, it has been shown that the inevitable
small remaining differences in social desirability between the components
of forced-choice items become magnified in this context, thus negating
the point of the procedure (Corah et al., 1958). Again, it must be said that
social desirability can be largely eliminated in careful test construction and
validation.
The tendency to endorse extremes Some subjects tend to endorse

extremes, when rating scale items are used. This is difficult to avoid but
again careful item writing and test construction can minimise the effects as
Guilford (1959) has argued.
Conclusions
It can be concluded that these response sets can be largely nullified by
careful item writing and even more important by careful test construction
and validation as will be discussed later in this chapter.
The other main difficulty with penonality questionnaires concerns the
ease with which subjects can deliberately distort their results, simply
because it is usually easy to see the point of questionnaire items. This is a
serious defect if questionnaires are used in a selection procedure. One way
to obviate the problem is to include a lie scale consisting of items on which
it is easy to detect distorten. The results of subjects who score beyond a
certain point on such a scale are ignored. Typical lie scale items are: I have
never told a lie; I always keep promises; I always hand in any money which
I find.
Another method which seems to work well, as judged by reduction of
lie scale scores, is to announce that cheating can easily be detected. Of
course, in selection, to be branded a cheat is a considerable threat.
In brief, it can be seen that deliberate distortion can be reduced but,
obviously, a test which was not so open is certainly to be preferred in
selection.
The advantages ofpersonality inventories

The strength of personality questionnaires is that it is easy by virtue of
methods of test construction to make them reliable, at least to determine
their validity, to make them discriminating and to establish good norms.
Thus the best personality questionnaires do possess the characteristics of
good test, discussed earlier in this chapter.
In addition, personality questionnaires are simple to administer and
score, especially if presented on a computer as is easily possible, and can be
given to large numbers of subjects at once. This makes them well suited
to applied psychology.
Since it is clear that many of the advantages of questionnaires come
about from methods of test construction these must be briefly described.
The constrnction ofpersonality questionnaires

I shall not describe in detail how personality questionnaires are constructed
because that would not be relevant to this chapter. In any case this can be
found in Kline (1986, 1992a). Here I shall delineate enough of the methods
for readers to understand how personality questionnaires can be made
reliable, discriminating and valid.
Item analysis in test construction The basis of test construction is to

administer items to samples of the population for whom the test is intended
and thus select items which have been shown to be efficient. In item
analysis, efficiency is judged against two criteria.
1 Items are selected if they are discriminating. The measure of this is that
not more than 80 per cent or less than 20 per cent put the keyed
response. Why this should be is obvious if we consider an item to which
all subjects had given the same response. Such an item gives no
information concerning individual differences.
2 Items are selected if they correlate beyond 0.3 with the total score. This
criterion ensures that all items are measuring the same variable which
is essential for a good test whose items are assumed to be from the same
universe of items. The higher the items correlate with the total score
the higher the internal consistency reliability.
Tests whose items meet these criteria are bound to be discriminating,
reliable and usually univariate, measuring one variable. This is important
since if tests measure two variables, as sometimes occurs, apparently
identical scores from two subjects may be psychologically different. Thus
a score of 10 might consist of 5 and 5 on the variables in the one case and
9 and 1 in the other. It should be noted that tests constructed by item
analysis, even if reliable and discriminating, still require validation, i.e. it is
necessary to demonstrate what the variable which the test measures actually
is.
Factor analysis in test construction This is highly similar to item

analysis and in most cases gives similar results. Thus items selected by one
method would also be selected by the other. Factor analysis, which lies at
the heart of psychometrics and is not simply used in test construction, is
fully discussed and explicated in Chapter 3.
Factor analysis is a statistical procedure which seeks to account for the
correlations between variables with as few dimensions as possible. Thus it
is ideally suited to test construction since the aim of the test constructor is
to produce a set of items whose correlations are accounted for by one
variable - the one the test is supposed to measure.
Thus in factor analytic test construction the items are given to a sample
ofsubjects and factored. All items which load (i.e. correlate with) the main
factor or factors, if more than one scale is being constructed, are selected
for the test. The other criterion of items being discriminating is also used.
Thus discriminating items which correlate with their requisite factors are
selected. Again, what the factor is, has to be demonstrated in studies of
validity.
The advantage of factor analysis over item analysis in test construction
is that it is able to select items which are truly univariate or unifactorial. If
a test were made up of items measuring two correlated factors item analysis
could produce a mixed test. In general, however, as has been stated, the
results of both methods yield similar results. Factor analysis, however,
demands larger samples and has certain technical difficulties, which are
often ignored and result in flawed tests, in the hands of poor test construc-
tors. All these problems will be discussed in Chapter 3.
There is a different method of test construction - the criterion-keyed
method- which was used for the MMPI, the most widely used personality
test in the world {Eysenck, 1989). In criterion-keyed test construction, a
pool of items is administered to different groups and items are selected if
they can discriminate one group from the others. Thus in the case of the
MMPI, items went into the test if they could discriminate among nine
clinical groups and controls. An item which could discriminate manics
wen~ into the manic scale, one which discriminated homosexuals into the
homosexual Scale and so on. Indeed, from the original MMPI item pool
more than 200 such scales have been developed by using it with other
groups (Dahlstrom and Walsh, 1960).
There are difficulties with this method oftest construction which render
it far from ideal. The main problem lies in the fact that groups may differ
on more than one variable so that a criterion-keyed scale is not necessarily
univariate and may measure a mixture of variables. Consequently that a
scale will discriminate among groups gives no precise indication of what
it measures. Thus criterion-keyed scales are empty of psychological
meaning. This makes theorising on the basis of their results of dubious
value. Indeed the use of such tests can, for this reason, hinder the
development of psychology. Furthermore, if groups are difficult to define,
as is the case in the clinical field, criterion-keyed tests may not discriminate
as well on subsequent use. Indeed, tests developed by this method are not
recommended except for screening purposes, for example in the armed
services where it might be necessary to exclude psychotics, and where the
only concern is that they are excluded.
Validating the test

By using item and factor analysis to construct personality questionnaires,
it has been shown that reliable and homogeneous tests can be produced.
However, it is still necessary to validate these tests. The validation of tests
has been discussed in the section on validity, and it is sufficient to repeat
here, that generally it is usual to demonstrate the construct validity of
personality questionnaires.
It is at this point oftest validation, however, that many tests are deficient.
For example, the PRF Oackson, 1974) is a superbly constructed test using
item analyses in a highly sophisticated fashion and rendering them as
efficient as factor analysis. However, the test variables were derived from
Murray's (1938) personological theory which has not been validated and
no effort was made by Jackson, in the test manual, to show that his scales
measured the needs postulated by Murray. Similar flaws can be found in
many of the personality tests used in social psychology (Robinson et al.,
1991).
Standardising the test

Test standardisation has already been discussed in this chapter. It is a simple
matter in principle to collect norms for personality questionnaires since the
tests yield clear unequivocal scores. The only problem concerns the time
and money required to collect sufficient samples.
Conclusions concerning personality inventories

From this discussion and description of penonality inventories it is clear
that they can be made reliable, valid, discriminating measures and with
good norms. Thus there is every reason to use them as basis for the scientific
study ofpenonality. However, I hope that it is also equally dear from the
discussion that penonality inventories per se are not necessarily good tests.
Indeed the majority of them are poor, unreliable and not valid, as is dear
from a perusal of any of the Mental Measurement Yearbooks (e.g. Buros,
1979) which contain reviews of current psychological tests.
Although some penonality inventories have the qualities required of
scientific measures and as such for the basis of the psychometric view of
penonality, there are other kinds of penonality test and these must be
described.
Projective tests
Projective tests consist usually of ambiguous stimuli which subjects are
required to describe. Their responses are then interpreted to provide
assessments ofpenonality. One of the most well-known psychological tests
whose fame has spread far beyond the boundaries of psychology is the
Rorschach test (Ronchach, 1921), a set of ten symmetrical inkblots. A
number of points need to be undentood before projective testing can be
evaluated.
-Projective tests are generally idiographic as distinct from penonality
questionnaires which are nomothetic. This means that projective
tests are concerned with what is unique to individuals and this is
often, as Allport (1937) has stressed, far more interesting than what
is common, and more salient to undentanding people. The fact that
they are idiographic has contributed, without doubt, to their popu-
larity among clinicians, whose work involves undentanding
individuals, and the same applies to many applied psychologists.
- The ambiguity of the stimuli. Stimuli are generally ambiguous and
visual. They are ambiguous because the description of a precise
photograph of, for example, a botde of Guinness, would be likely
only to provide a response 'a botde ofGuinness' except among the
severely psychotic. To recognise these, no test would be required.
The Ronchach test, as has already been mentioned, uses inkblots
as the stimuli. The TAT (Murray, 1938), or Thematic Apperception
Test, another famous projective measure, portrays human figures,
whose expression and even sex are unclear, in ambiguous contexts.

Many other projective measures use human beings, for example, the
Object Relations Technique (Phillipson, 1955), while those for
children sometimes substitute animals, on the grounds that children
will identify better with them (see below for the theory of projective
testing). Blum's Blacky Pictures (Blum, 1949) is a good example of
this.
-Other kinds of projective test stimuli. Not all projective tests make
use of ambiguous stimuli. The House Tree Person test (Buck, 1970),
as the name suggests, requires the subjects to draw a house, a tree
and a person and to answer certain questions about these drawings.
All these responses are then interpreted. Sentence completion tech-
niques on the other hand present incomplete sentences to subjects
who have to finish them. These endings are then interpreted. As
Semeonoff (1977) points out, in his excellent review of projective
tests, there are other types involving solid objects and even aural
stimuli but these are rare.
-Meaning of projection and the theory of projective testing. In
psychoanalytic theory (Freud, 1911) projection is an ego defence
mechanism in which unacceptable impulses are projected on to
others. For example, racists may see Black people as over-sexed and
aggressive. In projective test theory (Semeonoff, 1977) subjects
identify with the ambiguous figures or other stimuli and thus project
their own feelings and conflicts on to them. Thus if an inkblot is said
to show blood, that represents the aggression or the fear of the
respondent. Similarly if, in the TAT, a figure is described as terrified
this is supposed to represent the subject.
Two points should be noted: that this use of projection is quite
different from that ofpsychoanalysis and the assumption that subjects
identify with the main characters of the picture. There is nothing in
general psychological theory to suggest that either of these assump-
tions should be correct and this theoretical weakness of projective
testing (as Eysenck, 1959, has aq,rued) needs to be countered by
impressive empirical research demonstrating their validity.
-Reliability and validity of projective tests. The reliability and validity
of projective tests has been extensively scrutinised by many authors
(Eysenck, 1959; Vernon, 1964; Semeonoff, 1977; Kline, 1992a) and
there is little disagreement over the results. In general they are of
low reliability and validity. The low reliability is attributable to the
fact that almost all projective tests are subjectively scored. This gives
rise, inevitably, to inter-scorer unreliability and poor reliability ifthe
same scorer scores the test again. All this is bound to lower the
validity of the tests.
Eysenck (1959) has argued that the more rigorous the study of
the validity of projective tests the worse the results. He described
projective tests, indeed, as little more than vehicles for the rich
imagination of clinicians. Vernon (1964) showed that while it is
claimed that projective tests measure the inner depths of personality
(Murstein, 1963) in fact extraneous variables, such as the race of the
tester and her sex and their interaction with the race and sex of the
subjects, together with what subjects thought the tests were measur-
ing, all affected the results, findings which make it unlikely that deep
aspects of the personality are being measured. Indeed it must be
concluded that, as normally administered, there is little evidence for
the validity of projective tests.
-Normative data. Because of the nature of the responses, it is difficult,
but not impossible, to set up norms for projective tests. Exner (1986)
has gone a long way towards this for the Rorschach test, but ofcourse
this treats the data nomothetically. Even so, there are bound to be
responses which do not fit the norms.
Given all these difficulties the question must be answered as to why
psychologists persevere with projective tests. First, it must be admitted that,
in general, academic psychologists have abandoned them. Clinicians, on
the other hand, continue to use them for the following reasons. The
experience of administering projective tests is highly interesting. There is
a considerable variance in responses among subjects and these responses
appear to be psychologically meaningful and are often easy to interpret
along psychoanalytic lines. In a sense, they have high face validity. This
has to be considered together with the fact that there is a simplistic aspect
to personality inventories, as has been argued.
A second reason for not abandoning projective tests stems from the fact
that in the hands of skilled users it does appear that some remarkable results
have been achieved. One outstanding example, in this writer's opinion
(and this line of argument is inevitably subjective), is the work of Carstairs
(1957) with the Rorschach in his study of the Rajputs, data which would
have been unobtainable from questionnaires. Murray (1938) in his exten-
sive studies of personality was another brilliant exponent. In brief,
projective tests can be sources of remarkable data.
Even if this point were granted, however, it must be noted that a test
is not a satisfactory scientific instrument if it cannot be used by reasonably

intelligent and trained penonnel but requires some special gift or intuition.
This is certainly the case with projective tests. Thus we are left with a
dilemma: projective tests are sources of rich and unique data, yet clearly
on all examinations not valid or reliable, except, possibly, in the hands of
a few rare practitionen.
Scrutiny of the problems of projective tests, described earlier in this
section, give some hope that this dilemma may be resolved. Clearly much
of the difficulty sterns from the unreliability of the scoring procedures. If
these could be improved, it is possible that validity would also rise, although
it should be noted that reliable scoring is necessary but not sufficient for
validity. Work of this kind is already in progress. Thus Holtzman (1981)
has developed and researched a new form of the Ronchach, the HIT,
which utilises short answer questions to a large number of inkblots. This
is psychometrically efficient and seems to work well, although it appean
less rich than the original form of the test.
Another approach which has been attempted by Holley (1973) and by
Hampson and Kline (1977) is to score projective tests objectively. This
involves essentially a content analysis ofthe responses; one for the presence,
zero for the absence of a variable. An example will clarify the scoring
scheme. If a picture is described by one subject as having a rose in it, rose
is a variable and it scores one. All other subjects who put this score one.
Subjects who do not mention it score zero. In this way it is possible to
score most projective tests. The resulting matrix of ones and zeros can be
subjected to various forms of statistical analysis. Holley (1973) favoun G
analysis which involves a factor analysis ofthe correlations between subjects
rather than variables, thus clustering groups of subjects together. Holley
(1973) showed excellent discrimination with the Ronchach, by this
method, between depressives and schizophrenics. Hampson and Kline
(1977) also found that it would discriminate between various categories of
criminals, with the HTP test (Buck, 1970). However, there are various
technical problems with G analysis which are fully discussed in Kline
(1992a) which render its use more complex than the essence of the method
described here. Nevertheless it is a viable procedure for projective test
analysis although it turns them into nomothetic measures. In summary, this
is a possible method of combining the richness of projective tests with the
psychometric efficiency required of good testing.
Conclusions concerning projective tests

From this discussion it can be concluded that projective tests are not
satisfactory for the scientific study of personality, being neither reliable nor
valid, when administered according to their manuals. Objective scoring
schemes and methods of analysis, of which G analysis is only one example,
can certainly improve their psychometric efficiency but more research is
needed to ensure that by so doing, the unique richness of their data, which
is the reason for attempting to increase their efficiency, is not lost.
Objective tests
I shall now examine, rather more briefly, the third category of personality
tests- objective tests, defined by Cattell (1957), their main protagonist, as
tests which can be objectively scored and whose purpose cannot be guessed
by subjects.
In principle, any task which can be objectively scored and whose
purpose is impenetrable to subjects can be used as an objective test, given
also that there is variance in scores. However, this would lead to a virtual
infinity of tests so that in practice it is necessary in the development of
objective personality tests to draw on some rationale for the test from
experimental or clinical psychology. For example, it was noted in psycho-
analysis that delay in free association was evidence of the subject matter
touching upon emotional problems or conflicts. Jung (1910), indeed,
developed a test on this basis. Thus reaction times to words could be used
as an objective personality test. Cattell and Kline (1977) contains a full
discussion of principles of objective test construction.
Several points should be noted about this description of objective tests.
First it is clear that implicit in it is Cattell's concept of personality as the
totality of behaviour which involves understanding the determining fac-
tors. Secondly it is obvious that all such objective tests require validation.
Since, by definition, what they measure cannot be guessed, it is essential
that they have clear evidence of validity.
Examples of objective tests

Cattell and Warburton (1967) contains a description of more than 800
objective personality and motivation tests from which more than 2,000
personality variables have been derived. Most of these are still only at the
experimental stage and have relatively litde evidence for validity. Clearly,
in a chapter of this length, it would be impossible to illustrate even a small
proportion of them. Instead I shall simply give a few examples of objective
tests to indicate their range and scope.
1 The Fidgetometer. This is a chair with electrical contacts in the seat and
arms. This measures movements and is almost always undetected by
subjects. Even if subjects knew that they were being measured, it is hard
to fake. Is it better to be still or move a lot?
2 The Slow Line drawing test. Subjects are required to draw a line as
slowly as possible. Scores derived from this are the length of line and
whether subjects cheated or not by lifting their pencils above the paper
or stopped drawing.
3 Book titles. Subjects have to choose their preferred book titles. Selection
of socially acceptable rather than risque titles is claimed to load on a
measure ofassertiveness (Hundleby, 1973). The following tests also load
this factor.
4 Faster speed of tapping.
5 Faster speed of reading when asked to read at one's usual rate.
6 Greater preference for sophisticated activities.
Anxiety is measured by a number of objective tests, although the validity
of these measures is not as high as that of questionnaires.
7 Greater number of admissions of minor wrongdoings or frailties. It
should be noted that this is a questionnaire measure. However, it is an
objective test because no assumptions are made concerning the truth or
falsity of the responses. The measure is simply the number admitted to.
8 Greater acquiescence in answering questionnaires. Note here how what
in questionnaires is regarded as an annoying response set in objective
tests is used as a measure in its own right.
9 Addition under distraction. Scores on a simple arithmetic test are
obtained. Later in the test battery the same additions are presented with
jokes written on the sheet. Differences in score and time to complete
the additions are noted.
These tests give some idea of the nature of objective tests but a few more
general points can be made. Some objective tests are physiological. For
example, pupil dilation in the startle response (to a gun shot) is listed in
Cattell and Warburton (1967). All projective tests, if they are scored
objectively, are forms of objective test, and, as such, certain scores from
the Rorschach are listed in the Compendium (Cattell and Warburton,
1967). Questionnaires can be objective tests if responses of a certain kind,
e.g. the number of 'uncertain' or extreme responses, are counted regardless
of item content.
Few of these objective personality tests have been published as tests.
The two most well-known published batteries of objective tests are the
Objective Analytic Battery (Cattell and Schuerger, 1976) which measures

ten personality factors and the MAT (Cattell, Hom and Sweney, 1970),
measuring ten motivation factors.
I shall not describe or evaluate these tests here because the findings from
them will be discussed in Chapters 4 and 6. However, it should be pointed
out at this juncture that there are problems concerning their validity. For
example, a study of the MAT by Kline and Cooper (1982) found the scales
to be of low reliability and ten factors could not be extracted from the
items. Similarly ten factors could not be extracted from the OAB and the
factors which were extracted loaded, in some cases, on ability tests (Kline
and Cooper, 1984b).
Conclusions
The conclusions to be drawn from this discussion of objective tests are
clear. These are highly interesting and ingenious measurement devices but
are of virtually unknown validity. This is largely because they have been
little used beyond a small group of researchers working with or influenced
by Cattell. Most psychologists who are not experts in factor analysis have
been unable to penetrate the mysteries of the factor analytic arguments
supporting their use and their validity. Furthermore, because they lack face
validity, they have no direct appeal to testers. Combined with these
problems is the fact that they are not easy tests to administer, compared to
questionnaires, for example. Thus applied psychologists would have to
have clear evidence of validity and utility before they used them. All this
means that there is remarkably litde normative data for these tests which
further discourages their use. All this is sad for psychometrics because, in
principle, objective tests have many advantages over the other kinds of
personality tests.
COMPARISON OF THE THREE TYPES OF

PERSONAUTY TESTS
Personality questionnaires are simple to administer and score and have the
advantage that they are group administrable. In addition it is possible to
make them highly reliable, and to set up good norms. The best of them
have also been shown to be valid. This has meant that questionnaires are
widely used in the scientific study of personality and, above all, in applied
psychology.
On the other hand there is no doubt that personality questionnaires are
simplistic and it seems naive to think that personality can be fully concep-
tualised in terms of five variables (as are usually derived from question-
naires). On a purely practical point, personality questionnaires are far too
obvious and are easily distorted, which is a severe disadvantage as regards
their use in selection.
Projective tests, in contrast, capture much of the richness of personality
but do so at the expense of reliability and validity, except perhaps in the
hands of gifted or intuitive testers. Attempts to score projective tests
objectively have been made and deserve further research. However, until
the validity of such objective scoring schemes is established it is difficult to
use projective tests for substantive investigations of personality.
The same applies to objective tests. For these a great research pro-
gramme is required to discover what the huge array of objective tests
measure. This is worth doing because objective tests are difficult to fake,
overcome problems of response sets and are likely to be applicable
cross-culturally (certainly compared with many questionnaires). Their
objective scoring should ensure high reliability. Thus, in many respects,
objective tests are an ideal form of personality test. However, at present,
none is able to be used for substantive research.
Having described and evaluated the different types of personality test,
upon whose scores the psychometric view of personality is based, in the
next chapter I shall describe factor analysis and its application in personality
measurement. For it is through factor analysis that the results of personality
testing have been made useful both for personality theory and for applied
clinical, occupational and educational psychology, as is discussed in
Chapters 8 and 9.
Chapter 3
Factor analysis in the study of

personality
Factor analysis is a statistical technique central to the psychometrics of

personality. Not only have the best personality questionnaires been de-
veloped through factor analysis, but even more significantly, factor analysis
has been used to determine the most important variables in the field. This,
as was discussed in the opening chapter, is vital since there is little
agreement among personality theorists.
In this chapter my aim is not to explicate the algebra or computation
of factor analysis, of which excellent accounts may be found in Harman
(1976) or Cattell (1978) and clear explanations in Child (1991) or Kline
(1992a). Rather it is to enable readers to understand the logic of factor
analysis and to appreciate the problems which it can solve. In addition, I
shall set out the rules for technically adequate factor analyses because with
the ease of access of high speed computing in the social sciences, factor
analysis is used by researchers with little understanding of its underlying
mathematics, with the result that many reported factor analyses are mis-
leading. As shall be seen, this has been a considerable cause of error in the
psychometrics of personality.
Definition of tenns
First I shall define the terms used in factor analysis. Ishall begin with basic
statistical definitions since if these are not clear, the rest becomes hopelessly
muddled.
Variable
Any characteristic on which individuals, or one individual, over time can
vary is a variable. In the field of personality, anxiety or anger are variables.
Factor analysis 33
Variance
This is the variation in scores on a variable of a sample or population.
Sample and population must be distinguished. If we are studying anxiety
level among twenty-one- year-old males in Great Britain, the population
consists ofall such individuals. In most research we are forced to use samples
of populations and it is essential that samples are representative and
sufficiendy large to reduce statistical error. As shall be seen, sampling is
critical in factor analysis.
Correlation
The correlation coefficient r indicates the degree of agreement between
two sets ofscores. This correlation is positive ifhigh scores on one variable
are associated with high scores on the other. It is negative ifhigh scores on
one variable are associated with low scores on the other. The correlation
coefficient runs from +1 (perfect agreement) to - 1 (complete disagree-
ment), as would be the case if the order of scores on one variable were a
perfect inversion of the scores on the other. A zero correlation indicates
that there is no relationship between the two sets of scores.
Since correlations are the basis of most factor analyses in the field of
personality it is necessary to say more about them.
Meaning of correlations The square of a correlation coefficient

indicates the amount of variance in common between the two sets of
scores. For example, a correlation of0.7 shows that there is 49 per cent of
variance in common between the two variables. However, before this
figure becomes meaningful the determinants of correlations must be
discussed.
One cause of correlations between two variables may be elements
common to both. For example, there is a correlation between dogmatism,
as measured by Rokeach (1960), and obsessionality. This is because the
obsessional personality shares certain characteristics with dogmatism, and
some items in measures of these variables are highly similar.
In other instances, however, a cause common to both variables may
determine the correlation. There is one outstanding example of this in the
field of personality. There is litde doubt that there is a correlation between
cigarette smoking and lung cancer. Eysenck (1980) however has argued
not that smoking causes lung cancer but that the correlation is accounted
for by a third factor which causes an individual to smoke and to be liable
to lung cancer. This is a perfecdy possible interpretation of the correlation
34 Factor analysis
and it highlights a critical point about correlations: correlations do not

imply causation. Since the war there is a correlation between increases in
the mean height of adults and increases in divorce rate. Even sociologists
have not argued a causal connection.
Magnitude of correlations The size of the correlation indicates how

much variance is in common between two sets of scores, as has been
argued. However, this can be distorted by various factors which should
always be borne in mind in interpreting correlation coefficients. Since
factor analyses are usually based upon correlations, these same distortions
can affect factor analyses which is why these must be discussed.
Homogeneity of variance reduces the magnitude of the correlation
coefficient. There is a small negative correlation between extraversion and
academic success at the university, just for example, presumably because
extraverted individuals do not enjoy the lack of activity and concentration
required of academic work. However, this small correlation can only be
computed within students, i.e. subjects who have been able to do well
enough academically to enter the university. If we were to sample the
whole range ofextraversion (including those so extraverted that they could
not attain university standards) and the whole range of academic ability,
the correlation would probably be larger. The relationship between
homogeneity of variance and the size of correlations can be seen if we
consider the case where all subjects had the same score on extraversion.
Here the relation of extraversion to any other variable would be bound to
beO.
In many real life studies (as distinct from illustrative examples) samples
are restricted in variance, especially ability variance, and this restricts the
size of correlations. There are correcting factors which may be applied to
estimate the population coefficient from the sample, but in this writer's
view, it is better to sample efficiently and ensure a representative correlation
coefficient.
The other important distorting factor in correlations is the unreliability
of the measures. The lower the reliability the lower the resulting correla-
tion between them. As has been discussed in the previous chapter, any
decent test should have a reliability of at least 0.7. However, many
personality tests do not reach even this modest criterion and, in conse-
quence, correlations with such scales are lower than they would be if the
tests were perfectly reliable. Again there are formulae for correcting the
attenuation of correlations due to unreliability but again, in this writer's
view, it is better to use reliable tests than to adjust obtained statistics. As
Factor analysis 35
Cronbach (1976) has argued, poor data are always poor data no matter
what one does with the figures.
One of the aims offactor analysis is to account mathematically for a set
of correlations in terms of a smaller number of variables. Therefore, the
more accurately these correlations reflect the population variance and are
not artifacts of the homogeneity of the variance in the samples or the
unreliability of the tests, then the better the factor analysis will be.
Factor analysis
Factor analysis is a statistical method in which variations in scores on a
number ofvariables are expressed in a smaller number ofdimensions. These
dimensions are the factors. In the majority of psychometric studies of
personality, and indeed in all fields in which psychometrics is used, factor
analysis is applied to the correlations between variables. It mathematically
accounts for these correlations in terms of a smaller number of factors. The
factor analysis computes the correlations of each of the variables with these
factors. These are the factor loadings which define the factors.
I shall clarify this general description with a simple example, from the
field of personality. Suppose that we had obtained ratings ofa large number
of subjects on a variety of personality traits, e.g. talkativeness, sociability,
noisiness, trepidation, pessimism, tension and so on, and computed the
correlations between them. If we had ratings on about 100 variables it
would be impossible to work out in one's head any pattern of correlations.
Factor analysis is an ideal analytic technique because it attempts to give a
mathematical account of the matrix of intercorrelations, with a smaller
number of variables than the original set (reducing ranks of the matrix) . A
matrix is simply a set of numbers arranged in rows and columns, in this
case the correlations between the ratings.
In studies of this kind there are often five-factors, the big five referred
to in the opening chapter (Digman, 1990), of which two are particularly
important. One might well load on the following variables: sociability,
noisiness, friendliness, conviviality, energy; the second would load on
timidity, fearfulness, feeling sick before big events, poor sleeping, for
example.
It has been argued that factors are defined by their loadings: the
correlations with the variables. Thus to identify the first factor we have to
think what construct in the field of personality fits the description -
friendly, noisy, etc. In fact, extraversion as described by both Jung and
Eysenck fits well. Similarly the second dimension can be identified as
anxiety. In other words, the first two factors, anxiety and extraversion,
36 Factor analysis
account for the observed correlations between the variables loading on the
factors. As has been mentioned, the factor analysis of personality
questionnaires and ratings can usually be reduced to five factors. This
means that much of the variance in personality is explicable in terms of
five constructs. These would appear to be the most important variables.
In psychometric .studies of personality, and the results of these studies
will form the bulk of the remaining chapters of this book, it is not sufficient
to rely on the factor loadings alone to identify factors. Further experimental
work is required to validate the identification.
This schematic example of the factor analysis of personality ratings
illustrates clearly how complex data can be simplified by factor analysis. It
is impossible for most human beings to hold in their minds the correlations
between 100 variables, or even just the 100 variables themselves. For
example in attempting to assess what horse might win the Derby there are
far too many variables to be considered for most to make a correct choice.
What happens is that people select out what they believe are the most
important variables and consider those. Factor analysis puts this subjective
selection procedure on to an objective statistical basis. In the example from
personality, these five factors embrace much of the variance from a huge
number of variables. As shall be argued, the statistical basis for selecting
variables is superior to the intuitive judgements of personality theorists
among whom there is no agreement.
With this preliminary discussion of factor analysis I shall now define
some more specific terms and say a little more about factor analysis in
general.
Exploratory factor analysis In the illustration of factor analysis in the

previous section, it was used to simplify a large set of data, to identify the
most important variables. For obvious reasons this is called exploratory
analysis. This is much the most common use of factor analysis in the study
of personality and in psychometrics generally. However, there are other
kinds of analysis.
Confinnatory factor analysis In this method factor analysis is used to

confirm hypotheses. Thus, to take the example of the big five, it would be
possible to obtain a further set of personality ratings and to investigate
whether the data fitted the five-factor hypothesis by using confirmatory
factor analysis.
In this technique the factor loadings are hypothesised and the compu-
tations involve fitting the target matrix as closely as possible. It is possible
to test the fit statistically and modem factor analysts prefer confirmatory
Factor analysis 37
analysis because the factors can be confirmed statistically. Nevertheless

there are difficulties in its use which will be discussed later in this chapter.
Definition of a factor As Nunnally (1978) points out, strictly a factor

is simply a linear combination of variables. These combinations can involve
any kind of weightings of the variables - differential, negative, unit, or
positive. Factor analytic procedures are methods of calculating these
combinations and weights, with the aim of explaining the most variance
with the fewest factors, at least in exploratory analysis.
This is the statistical definition. However, most psychologists defme a
factor as a construct operationally defined by its factor loadings (Royce,
1963).
Factor loadings As has been seen, factor loadings are the correlations
of variables with the factors, although in oblique rotations (described
below) this is true only of the structure matrix. The importance of factor
loadings is tied in with the definition of factors. Thus while everyday
definitions of extraversion are bound to vary, the definitions of the
extraversion factor are precise: it is defined by loadings on given variables.
There is a further advantage to factor analytically defined concepts
compared with normal concepts, over and above their precision. A normal
(by which I mean developed through thought or reasoning) concept may
be of little value. For example, phlogiston, although ingenious, in fact
explains nothing. However, a factor explains variance and, if a large factor,
explains a considerable amount. It is, therefore, necessarily a useful
concept.
Factor identification It was argued above that factors are identified

from their loadings but that identification should be extended to other
criteria. There are various reasons for this. Thus in many cases identification
from loadings alone is not unequivocal and is not as clear as appears in the
examples. Furthermore some authors who are opposed to factor analysis,
e.g. Heim (1975), have argued that factors are simply statistical entities
which refer to nothing beyond the correlation matrix: an argument which
they buttress with the fact that there are many different factor analytic
solutions to the same correlation matrix. Consequently, any factor needs
external verification.
Certainly, it is good psychometric practice to verify factors experimen-
38 Factor analysis
tally and almost all the factors which will be discussed throughout this book
have been thus identified. Heim's point about the multiplicity of factor
analytic solutions will be discussed later in this chapter.
Communality This is the proportion of variance of each variable

accounted for by the factors. It is important to know how well a factor
analysis accounts for the variance in a matrix of correlations. There are
several criteria to judge the adequacy of factor analyses of which one is to
examine the communality of the variables. To calculate the communality
of each variable, the common variance accounted for by the factors, the
factor loadings of a variable on each factor are squared and added. Thus
for example, if in a three-factor solution, a test of obsessional personality
loaded 0.71 on factor 1, 0.02 on factor 2 and 0.10 on factor 3, these factors
account for just over 50 per cent of the variance on the test. This is not a
particularly good solution since nearly half its variance is unexplained. This
is called the unique variance which consists of variance unique to the test
and error variance and is of little psychological interest. In another study
with different variables and factors the communality could well change.
Thus it can be seen that one demonstration of the adequacy of a factor
analysis is that the communalities are large. Ideally more than 80 per cent
of the variance of each variable should be explained.
Reproducing the correlations

Another measure of the quality of a factor analytic solution is to see how
accurately the original correlations between the variables can be repro-
duced. Simple cross multiplication of the loadings (see Cattell (1978) for a
worked example) yields the correlations and the closer to the originals these
are the better. The fact that the correlations can be reproduced from a
smaller number of factors is another indication of the power of factor
analysis in understanding the variance in a correlation matrix.
Eigen values or latent roots A third test of a factor analysis is to see

how much variance overall the factor explains. The eigen values or latent
roots indicate how much variance each factor accounts for and are
calculated by summing the squares of each loading on a factor.
li,
Factor analysis 39
'F
I
I
----- a, =. 5 . •.. :
I
~ ······."!-· a, = .29
a2 = - .57 1
-a-2-=--~.4~~~0~--------F 2
Figure 1 Orthogonal rotation of factor loadings

Source : Adapted from Cattell (1978)
Factor rotation Factors can be conceptualised as axes, as shown in

Figure 1.
From this it is clear that factor loadings change, depending upon the
position of the axes, although each position is mathematically equivalent.
Since factor loadings define factors it is obvious that where rotation should
stop is critical to the whole factor analytic procedure. Indeed the fact that
there is an infinity of mathematically equivalent factor solutions creates
severe problems which must now be scrutinised.
Problems in factor analysis

The essence of the difficulty in factor analysis is simple. On what grounds
is one solution from this infinity of possibilities to be chosen? More
technically, where does rotation stop? Since there may be no reason, a
priori, to choose one rotation than another it has led some psychologists,
notably Heim (1975), to argue that the method is not valuable.
Indeed, a superficial examination of the results of factor analysis in the
field of ability (see Kline, 1991 for a summary), and in the field of
personality, would lead one to a similar conclusion for factor analysts appear
to propose many different and disparate sets of factors . However, modem
factor analysts have overcome all these problems and replicable and
meaningful solutions are obtainable. How this can be done and the reasons
for these disagreements among different analysts must now be scrutinised.
There are both logical and technical difficulties and these will be dealt with
separately.
40 Factor analysis
Logical difficulties
One cause of disagreement concerns the identification of factors. There is
a clear example of this in the work ofCattell (1957) and Eysenck (Eysenck
and Eysenck, 1975), which is discussed in Chapter 4. Both agree that there
is a large factor loading on variables relating to psychiatric disturbance and
neurotic disorders. Furthermore, they have been shown to be the same
factor in empirical studies (Kline and Barrett, 1983). However, for reasons
bound up with their own theories, Cattell has labelled the factor 'anxiety'
and Eysenck 'neuroticism'. However, this difference of nomenclature is
unimportant from the viewpoint of understanding the structure of person-
ality and for its application in applied psychology.
Simple structure
However, as was made clear in the previous section the essential problem
of factor analysis - the choice of solution from the infinite possibilities - is
more complicated. Thurstone (1947) supplied a logical answer, which still
constitutes the principle behind the modern solution to this difficulty.
Each solution may be regarded as a hypothesis to explain the correlation
matrix. In the scientific method there is general agreement that if there are
competing hypotheses, the simplest is to be preferred - the law of
parsimony or Occam's razor. Thurstone, therefore, aimed to rotate the
axes to simple structure, this being defined as the most simple position.
The essence of rotation to simple structure is to arrive at a solution which
maximises the number of zero or near zero loadings. This ensures simple
factors each with a few high loadings.
The elegant logic of Thurstone's arguments has received empirical
support from the work of Cattell (1978) who has shown that simple
structure solutions are replicable and yield meaningful factors in cases
where the factors are known. Thus simple structure rotations go a long
way to answer the problem of the infinity of solutions.
Unfortunately, as Cattell (1978) has demonstrated, although most of the
leading psychometrists agree that simple structure should be the aim of
factor analysis, there is little agreement on how simple structure may be
obtained. Furthermore, Cattell (Cattell, 1978; Cattell and Kline, 1977) has
shown that simple structure cannot be attained unless the factor analysis
satisfies certain technical criteria. This is particularly serious because it can
also be shown that many published factor analyses are technically flawed.
Indeed Cattell (1978) argues that it is these technical errors which have led
to the disparity of results. Thus it is now necessary to set out the technical
Factor analysis 41
demands of good factor analyses and to discuss the concomitants of the

many different types of technical error.
Technical difficulties in factor analysis
Sampling variables
In exploratory factor analyses, as has been discussed, the aim is to map out
the field. To do this it is essential that all variables are sampled. For example,
ifno measures ofanxiety were included in a study there could be no anxiety
factor. A weakness of many studies is that the selection of variables is
arbitrary.
Furthermore a factor requires to be marked by at least three variables
which means that one test of a variable is insufficient. Clearly then it is
critical that variables be properly sampled in exploratory factor analysis.
Sampling subjects
This is important in factor analysis because if the variance on variables is
restricted, factors cannot emerge with any clarity because, as has been
argued, correlations become attenuated. This can be important in the field
of personality where, for example in normal subjects, abnormal personality
factors have low variance.
Numbers of obseMJations and variables

For reasons of matrix algebra it is essential that the number of subjects
exceeds the number of variables. If it does not the results are essentially
meaningless, as has been shown by Nunnally (1978). However, the number
of subjects does affect results as does the ratio of subjects to variables.
Number of subjects
This can be quickly dealt with. The larger the N the better because this
reduces the standard errors of the correlations. Where N drops below 100
results should be treated with caution and replication is essential. It is
noteworthy in social psychology and health psychology that, where factor
analysis is frequently carried out by psychologists who depend entirely on
computer programs, Ns fall well below this figure.
42 Factor analysis
Subjects: variables ratio

Here, until recently, there was considerable disagreement among the
leading authorities. Nunnally (1978) was the most conservative, claiming
a ratio of 10:1 was necessary, while Guilford (1958) was at the other
extreme, with a ratio of2:1.
Barrett and Kline (1981) noted that there was no rationale for any of
these ratios other than the experience of the writen. They carried out
therefore an empirical examination of this issue in which items from
Eysenck's EPQ and Cattell's 16PF Test (both of which will be discussed
in some detail in the next chapter) were subjected to factor analysis. With
a subjects: variables ratio of20: 1, a factor structure of such clarity emerged
that it was used as a baseline with which to compare the results from other
smaller samples. It was shown that with a ratio of 2:1 the main facton
emerged and that with 3:1 differences from the large sample were trivial.
These findings were essentially confirmed by Arrindel and Ende (1985)
who also showed that the ratio of subjects to facton extracted was critical.
In conclusion, however, a factor analysis with at least 100 subjects and a
subjects:variables ratio of at least 3:1 ought to be sound in respect of
statistical error.
Principal components or principal factors

Computing packages all offer the choice of principal components or
principal facton and which is preferable must be briefly discussed. Fint it
should be noted that components analysis is not strictly a form of factor
analysis because it yields as many facton as variables and explains completely
the variance in the particular matrix. Facton on the other hand are
hypothetical because they are estimated from the correlation matrix. This
makes it more likely, in principle, that they could be generalised to other
matrices than could components based on a particular matrix. This is a
theoretical point in favour of principal facton.
In computing principal components all the variance in the matrix,
including error variance, is explained. Principal facton, by estimating
communalities of the variables, attempt to exclude the error variance and
Carroll (1983) has argued that this makes principal factors a more accurate
procedure. However, Harman (1976) has shown that in large matrices of
more than 25 variables there is virtually no difference between the facton.
In this writer's opinion, the distinction is oflittle practical significance.
I have usually computed both methods and the differences are indeed, for
the practical purpose of identifying meaningful facton, trivial. Given the
Factor analysis 43
faint possibility ofinaccuracy in components analysis it would seem sensible

to compute principal factors.
Maximum likelihood factor analysis

In recent years this method of initial factoring (based on the work of
Joreskog, 1973) before rotation has become popular because programs are
now available for the formidable computing involved. This method obtains
a set of factors which each in tum explains as much variance as possible in
the population matrix, as estimated from the sample correlation matrix.
This is, therefore, an inferential statistical method because it seeks to
extrapolate from sample to population. Principal components, of course,
apply strictly only to the sample matrix.
Maximum likelihood analysis can be used in exploratory factor analysis.
If it is, however, rotation is required to give meaning to the factors.
However, when communalities and test reliabilities are high the differences
between principal components analysis, principal factor analysis and maxi-
mum likelihood analysis are small, as was found by Kline and Lapham
(1990) in the development of the PPQ.
The claimed advantage of maximum likelihood analysis in exploratory
factor analysis arises from the fact that there is a statistical test for the
extraction of each factor, whereas other methods of factor analysis are
essentially algorithms. However, this statistical test for the number offactors
is not always sensitive enough in large matrices to select between various
options, thus negating the advantages of its statistical basis.
It may be concluded that for exploratory analyses maximum likelihood
methods have little or no advantages over other procedures. However,
their main use is in confirmatory analysis and this will be discussed later in
this chapter.
Rotating the correa number ofjaaors

Cattell (1978) argues that this is one of the critical issues in reaching simple
structure because if too many factors are rotated factors split. If too few
factors are rotated a few broad factors are produced, dimensionality being
compressed.
Kline and Barrett (1983) examined this problem in considerable detail
partly based on their own empirical studies of the problem (Barrett and
Kline, 1982). They concluded that there was no one single best method
for selecting the number of factors. However, the rotation of factors with
eigen values greater than 1, which is the default procedure for many
44 Factor analysis
statistical packages, was shown to be unsatisfactory in large matrices

considerably over-estimating the number of factors, as Cattell (1978) had
argued. Two tests seemed" to work well and to be in general agreement:
the Scree test (Cattell, 1966) and the Velicer test (Velicer, 1976). However,
there is a considerable subjective element in the Scree test which requires
some experience and practice with colleagues who can use it. To overcome
this Barrett and Kline (1982) developed an automated version, although
even in this there is a small subjective element. The Velicer test is quite
objective but there is less evidence in support of its validity.
In conclusion it seems best to use both these techniques and if they
suggest different numbers of factors both rotations should be made. These
methods are guidelines rather than rigid rules.
Orthogonal or oblique rotations
Orthogonal rotations In orthogonal rotations the factor axes are at

right angles to each other (as in Figure 1) and are uncorrelated.
Oblique rotations Here the factors are oblique, less than 90 degrees.
Oblique factors are correlated, the cosine ofthe angle between them giving
the correlation. It is obvious, if Figure 1 is borne in mind, that as the angle
becomes smaller the factors become more similar, until as they take up the
same position they become identical.
Thurstone, as has been argued, defined simple structure as a set of factors

each with a few high loadings but the majority zero or close to zero. With
such a definition it is obvious that oblique rotations should be better able
to reach simple structure than orthogonal rotations because their axes can
take up any position thus maximising the zero loadings. There is a further
argument in favour of oblique factors, namely that in the real world it is
unlikely that important determiners ofbehaviour such as factors would be
uncorrelated; points stressed by Cattell (1978) and Kline (1979). Thus
almost all factorists of any note support Cattell's position that oblique
rotations are essential for obtaining simple structure.
The only exception is Guilford (1959) who claims that while it is true
that an oblique factor may be more simple than its orthogonal counterpart,
a set of oblique factors is more complex than a set of orthogonal factors
because in interpreting them their correlations have to be taken into
account. This argument turns on the meaning of simple. However, in the
field of personality, as shall be seen, empirical work supports the notion
Factor analysis 45
that oblique sets of factors are more replicable and more simple. In any
case the argument that orthogonal factors are unlikely in the real world
should not be ignored.
There is a further reason why oblique rotations should be preferred. It
is possible to factor the correlations between oblique factors. These factors
are known as higher-order factors and are more broad than primary factors
which load on them. It is often possible to describe large matrices in terms
of a few higher-order factors and this is, indeed, a simple solution which
is only possible with oblique factors.
In summary, therefore, it must be concluded that oblique rotations are
essential if simple structure is to be obtained.
Methods of rotation
Cattell (1973, 1978) reiterates the point with which this writer is in
complete agreement that one of the major causes of failure to reach simple
structure is poor rotational methods, even when these produce oblique
factors. This is a difficult problem because there are now many different
oblique rotation programs, although there is some agreement as to the best
procedures. I shall summarise the conclusions, although for a more full
discussion readers should see Kline (1992a) and Gorsuch (197 4) for
computational details, because if simple structure has not been obtained,
the results of studies, as will be seen in subsequent chapters of this book,
must be discounted.
Hakstian (1971) studied a number of oblique rotational procedures and
found that the most efficient and best at obtaining simple structure was
Direct Oblimin Qennrich and Sampson, 1966), while the computationally
more simple Promax was also good, provided that the factors were not too
oblique. If for some reason orthogonal rotation is desired, Varimax is
without question the best. There is surprisingly wide agreement concern-
ing the effectiveness of these methods although Cattell (1978) argues that
his own technique Maxplane followed by Rotoplot (in which the factors
are hand adjusted) yield the best results. However, this method requires
no little skill, and as Hakstian (1971) found, it is generally no better than
and sometimes less effective than Direct Oblimin.
These rotational procedures try to attain simple structure by maximising
the hyperplane count. Hyperplanes are boundaries alongside factors. If
these are fixed at+ or-o.OS (regarding all loadings within these boundaries
as zero} then the hyperplane count is the number ofloadings within these
hyperplanes. Maximising this maximises the number of zero loadings
which is an important criterion of simple structure.
46 Factor analysis
The technically sound factor analysis

From this discussion the technically sound factor analysis can now be
described. There will be proper sampling of variables and subjects. After
principal factor analysis (or possibly maximum likelihood analysis}, the
significant facton as selected by the Scree or Velicer tests (or possibly from
the maximum likelihood analysis) will be rotated using Direct Oblimin.
These simple structure facton will be replicable and should ;lccount for
much of the variance in the matrix.
In addition to the fact that simple structure facton are replicable and
conform to the notion of panimony, as has been argued, Cattell (1978)
contends that they are causal determinen. This argument rests on two
points. If random data are factored, simple structure, as defined by the
hyperplane count, cannot be obtained. More importantly if artificial data
sets with known determinen are factored, simple structure analysis reveals
them, as is exemplified in the study of the behaviour ofballs. Here three
facton emerged: weight, size and elasticity. This makes it clear that simple
structure facton are not algebraic artifacts.
Now that the characteristics ofgood factor analyses have been described,
it will be possible in the subsequent chapten of this book to scrutinise
critically the psychometric studies of personality which appear to have
produced a plethora of contradictory findings. Much of this research will
be found wanting, not only with reference to inadequate factor analyses
but also in respect of poor tests, as discussed in Chapter 2, and thus results
may be ignored.
However, before we tum to an examination of the substantive work in
the psychometrics of penonality a few other points about factor analyses
should be made.
Factor structure and factor pattern

In oblique analyses a factor pattern and a factor structure are produced
which in almost all cases are highly similar. In orthogonal rotations, pattern
and structure are identical. Structure loadings are the correlations of the
variables with the facton. This is the definition of factor loadings given
earlier in this chapter. The pattern loadings are beta weights indicating the
importance ofeach variable in predicting the facton. The structure loadings
are more stable from study to study and the factor structure should be the
basis of the interpretation in oblique analyses.
Confirmatory factor analyses Most factor analyses in the field of
Factor analysis 47
personality are exploratory analyses. However, on occasions it is desirable

to test hypotheses, and for this confirmatory factor analysis is used. For this
there are two possible methods and these will be briefly described.
Procrustes rotations to target matrices In this method a target

matrix of loadings is specified on the basis of theory or previous findings
and a Procrustes analysis is computed which aims not at simple structure
but to match this target matrix as closely as possible. The more precisely
the target matrix is specified the more difficult it is to fit, the loadings of
simply zero, positive or negative are regarded as an easy target. Guilford
(1967) has been the leading exponent of this technique in his studies of
human abilities.
This method ofhypothesis testing cannot now be supported since Hom
and Knapp (1973) showed that target rejection was most unlikely unless
the target matrix was specified in detail. They further demonstrated that
Procrustes could match target matrices with random data and with data in
which hypotheses antithetical to the target were built in. Since, in addition,
simple structure factors are preferred because they are simple, the status of
Procrustes factors, even if the technique could be trusted, is dubious. Much
depends here on the theory from which the target matrix was derived, a
criticism which applies equally to the maximum likelihood method dis-
cussed below.
Maximum likelihood confirmatory analysis This is the favoured

method for testing hypotheses through factor analysis since there are
statistical tests of fit . However, as Nunnally (1978) has pointed out, the
chi-square test of fit finds it difficult to choose between target matrices
unless these are grossly different with the result that judgement still enters
the procedure.
However, despite its statistical sophistication, confirmatory analysis has
yet to yield any powerful substantive findings in the study of personality
beyond those from simple structure analyses. Ifcomputing time is available,
both methods deserve scrutiny.
R technique Most factor analyses are computed on the correlations

between variables. This is known as R technique. However, Cattell (1978)
describes some other approaches of which the most important for the study
of personality are set out below.
P technique Here test scores obtained from one individual are factored.
48 Factor analysis
This can be powerful in the study of motivation and of variables which

change over time.
Q technique Here the correlations between people are factored; the

factors thereby revealing groups. This is a useful method in clinical
psychology for example, where groups might be found differing on critical
personality variables.
0 technique Here scores from subjects on two occasions are factored.

This can be useful in the study of ongoing processes such as psychotherapy
or educational methods.
Conclusions
Sufficient has been said about factor analysis to enable readers to follow the
evaluations and criticisms of research which will occur throughout the
remainder of this book. The critical issue is that of simple structure. Simple
structure factors are replicable and are the logical choice from the infinity
of possible factors. That they can be causal is also important. They are
attained only in technically adequate factor analyses which involve proper
sampling of variables and subjects, and the oblique rotation (by Direct
Oblimin) of the correct number of factors chosen by the Scree or Velicer
tests. Maximum likelihood methods were permissible but their statistical
claims should be treated cautiously. Similarly confirmatory factor analyses
need careful scrutiny, for their statistics are not highly sensitive.
Chapter 4
The factor analysis of

temperament
In this chapter I shall summarise the main findings from the factor analysis
ofpersonality questionnaires which is essentially the study of temperament.
As should be obvious from Chapters 1 and 3 of this book, there are many
and diverse findings both on account of the difficulties of defining what is
meant by personality and the problems of factor analysis. In a chapter of
this length a careful description of all the factor analytic research would be
pointless, since much ofit fails to meet even the least stringent of the criteria
discussed in the previous chapter.
I shall discuss, therefore, what is generally regarded, in terms of citation
and reference, as the best work in the field, most of which as shall be seen
is technically sound, even though in some cases, more recent research has
shown the results to be not the best factor analytic description of person-
ality.
I shall begin with the work of the great pioneers in this field - Guilford,
Cattell and Eysenck - who have produced factors whose psychological
meaning is well explicated by research. I shall deal then more briefly with
some other factor analytic sets of factors and the work on the 'big five'
which were mentioned in the previous chapter. Finally, based on this
discussion I shall delineate the factor analytic picture of human tempera-
ment, as it appears in 1992.
The work of Guilford

Guilford et al. (1976) in their Handbook to the Guilford-Zimmerman
Temperament Survey summarise much of the research with the Guilford
factors and this most recent account forms the basis of our discussion.
However, it should be pointed out that this research was among the earliest
factor analytic work in the field of personality much of which is described
in detail in Guilford (1959). The Guilford factors are set out below.
50 The factor analysis of temperament
G- General activity: energetic, quick vs. slow, deliberate.

R - Restraint: serious minded vs. impulsive.
A - Ascendance: assertive, confident vs. submissive, hesitant.
S- Sociability: friendly, talkative vs. shy, withdrawn.
E- Emotional stability: cheerful, composed vs. gloomy, excitable.
0 - Objectivity: tough vs. tender-minded.
F- Friendliness: respect for others vs. hostility, restraint.
T - Thoughtfulness: reflectiveness vs. interest in the outer world.
P - Personal relations: tolerance of people vs. fault-finding.
M- Masculinity: hardboiled, emotionally inexpressive vs. sympathetic,
emotional.
There are several important points to be noted about these factors which
will be salient to our discussion throughout this chapter and which will be
set out below. Before this is done, however, the status of these factors must
be emphasised. These ten factors represent, according to Guilford, the ten
most important variables in the personality field. It is these factors which
should be considered in trying to understand the nature of personality, and
develop any theory about its development and origins. The bipolarity of
the factors indicates the meaning of the low and high score on the variables.
1 Reliability of the scales. Against the criteria of good tests discussed in
Chapter 2 these scales are excellent; internal consistencies are mostly
beyond 0.8 and test-retest reliabilities after one year are also high.
2 Certain variables in this list are noteworthy: tough-mindedness; emo-
tional stability; sociability and thoughtfulness. These are common to the
work of most factor analysts. It is particularly interesting that sociability
and thoughtfulness are separate factors.
3 These factors are orthogonal. The argument concerning the importance
of simple structure and the fact that oblique factors were more likely to
be simple was discussed in Chapter 3.
4 Validity of the factors. As was pointed out in Chapter 2, it is difficult
to demonstrate the validity of personality factors except through studies
ofconstruct validity which necessarily involve the consideration oflarge
numbers of investigations. Guilford et al. (1976) summarise a consider-
able body of research in the test handbook but, as has been pointed out
by Kline and Barrett (1983) and Kline (1992a), many of the studies cited
there have samples too small for reliable conclusions to be drawn.
However, there are a few critical studies which illuminate the psycho-
logical meaning and scientific status of the Guilford factors and these
will be discussed.
The factor analysis of temperament 51
Eysenck and Eysenck (1965), in a joint study of the Guilford, Cattell and
Eysenck scales (a research which will be discussed later in this chapter when
the Cattell and Eysenck scales are scrutinised), attempted to locate the scales
in factor space, an efficient method of investigating the validity of factor
analytic tests. These authors carried out an oblique (Promax) rotated factor
analysis of items in the Guilford, Cattell and Eysenck scales. However, the
Guilford factors did not emerge and it was concluded that these factors
were not a good account of personality, probably due to the fact of the
orthogonal rotation.
However, only eight items per Guilford scale were used (to reduce the
size of the computations) which, as has been shown in Chapter 2, must
contribute to lowered reliability of the scales. Furthermore, the Promax
rotation may not have reached simple structure (there was no Direct
Oblimin at this time) . Thus this research, although an impressive contribu-
tion, was less than definitive through technical deficiencies of the
computing facilities when it was undertaken.
Cattell and Gibbons (1968) administered items from the Guilford and
Cattell scales to a large sample of students and submitted them to a factor
analysis which, as might be expected from these authors, conformed fully
to the criteria of technically adequate factor analyses. In addition they
performed an orthogonal rotation. They found that the Guilford factors
were essentially identical to those of Cattell when rotated to the oblique
position, the majority aligning with the Cattell factors although some were
· a mixture. Since they could not replicate the Guilford factors orthogonally,
they concluded that the Guilford set were no different from those of Cattell
but were improperly rotated. However, this conclusion must be discussed
again when the Cattell factors are examined, since there is considerable
doubt about their structure (see Kline, 1992a).
A study by Amelang and Borkenau (1982) must be mentioned. More
will be said about the relationship of the big five factors to other sets of
factors at the end of this chapter.
Conclusions concerning the Guilford set offactors

These factors are among the most interesting personality factors since they
have their origins in the first factor analyses of personality. However,
despite the attractiveness of some of the variables, in terms of common
human experience (see G and M, for example) and the fact that the scales
are reliable, it is difficult to argue that these factors offer the best possible
factor analytic description of temperament. The fact that they are ortho-
gonal is a priori somewhat unlikely. There is no personality theory which
suggests that the most important personality dimensions might be inde-

pendent of each oth~r. Furthermore even orthogonal rotations find it
difficult to locate the factors, whereas rotated to oblique positions, the
factors do not seem independent of other systems. In brief, a brilliant early
attempt to factor personality traits but not probably the most parsimonious
or meaningful factors that could be produced.
The Cattell facton

Cattell has been, without question, one ofthe most influential psychologists
in the world and since much of his enormous output has been concerned
with the factor analysis of personality, his findings must be scrutinised with
the utmost care. To summarise 40 or so books and about 550 papers and
chapters in one part of a chapter is impossible, but the essentials of his
empirical work can be set out. His theoretical account of personality which
is more complex will be discussed in Chapter 9.
This summary of Cattell is derived from a study of most of his work
but the most important single reference is Cattell (1981) which contains
an account of much of the empirical work and theory but is, by any
standards, a difficult book. Cattell and Kline (1977) is a more readable
discussion of his work and other important books and papers will be cited
at relevant points in the description.
The normal Cattell personality factors

These are measured in the 16PF Test (Cattell, Eber and Tatsuoko, 1970)
and in versions of the same test for adolescents (High School Personality
Questionnaire), primary school children (Children's Personality Question-
naire) and even pre-schoolers (The Pre-school Personality Quiz). Details
of all these tests can be found in Cattell (1973) and Kline (1992a). These
different tests allow longitudinal studies of the factors and research into the
influence of environmental factors on development.
As the name of the test implies, sixteen personality factors are claimed
by Cattell to account for much of the variance among personality traits
among adults, although there are seven other smaller factors (Cattell, 1973).
These sixteen factors are:
A- Reserved, detached vs. outgoing, warmhearted.
B- Crystallised intelligence.
C- Emotionally unstable vs. emotionally stable.
E- Humble, mild vs. assertive, dominant.
F- Sober, taciturn vs. happy-go-lucky, enthusiastic.

G - Expedient, disregards rules vs. conscientious, persistent.
H- Shy, timid vs. venturesome, uninhibited.
I -Tough-minded, self-reliant vs. tender-minded, sensitive.
L - Trusting vs. suspicious.
M - Practical vs. imaginative, bohemian.
N - Forthright, artless vs. shrewd, acute.
0 - Self-assured, secure vs. guilt-prone, apprehensive.
Q1- Conservative vs. radical.
Q2 - Group dependent vs. self-sufficient.
Q3 - Undisciplined, lax vs. controlled.
Q4 - Relaxed, tranquil vs. tense, frustrated.
These primary factors are oblique and at the second-order four factors are
extracted:
1 -Introversion vs. extraversion (exvia in Cattell's terminology).
2- Low anxiety vs. high anxiety.
3 - Sensitivity vs. tough poise.
4 - Dependence vs. independence.
Abnormalfactors
Before discussing this factor structure which is claimed by Cattell to set out
the major dimensions of personality and which forms the basis of his
theorising (Cattell, 1981), it should be pointed out that Cattell has also
identified a number of abnormal personality factors which are measured in
the Clinical Analysis Questionnaire (Krug, 1980). These factors are:
Dl. Hypochondriasis. 02. Suicidal depression. 03. Agitation.
04. Anxious depression. DS. Low energy depression.
06. Guilt and resentment. 07. Boredom and withdrawal.
Pa. Paranoia. Pp. Psychopathic deviation. Sc. Schizophrenia.
As. Psychasthenia. Ps. Psychological inadequacy.
Discussion of abnormal factors

Because these abnormal factors are a minor part of Cattell's work, I shall
discuss them briefly at this point before scrutinising the factor structure of
normal personality traits.
These factors are just about reliable enough to use with individuals,
having reliabilities around 0. 7. However, their validity is less well attested.
54 The factor analysis of t_
emperament
Kameoka (1986) examined the factor structure in a sample of students,

utilising an oblique Maxplane and Rotoplot analysis (see Chapter 3) as
advocated by Cattell (1978). He was able; to find the twelve factors although
the correlations among the depression factors were larger than expected.
It should be noted, however, that this study was among normals and thus
the variance would be restricted.
The CAQ is interesting in that it suggests that twelve factors can account
for abnormal personality. Furthermore the fact that there are seven de-
pression factors is obviously of interest to clinical and medical psychology.
Clearly these factors require much further research but they indicate how
factor analysis can illuminate complex fields of psychology.
Discussion <if the Cattell normal factors

The Cattell factors are so much more than factors derived from a set of
personality test items that it is necessary, if the claims of Cattell concerning
the importance of these factors are to be understood, to describe their
provenance and the external evidence for their psychological meaning.
Origins of the facton In Chapter 3 it was argued that in exploratory

factor analysis, proper sampling of variables was essential. In the develop-
ment of the 16PF Test, Cattell tried to ensure this by basing his items on
a previous analysis of ratings of personality. Originally Cattell searched the
dictionary for descriptions of personality, eliminated synonyms, and then
rated subjects on all remaining descriptors. These terms were said to
embrace the whole of personality- the semantic personality sphere -since
existence is dependent on descriptors. We live indeed in a linguistic world
(see Cattell, 1957 for a detailed description of this aspect of the research).
Factor analysis of these ratings had revealed twelve L (life) factors and
thus Cattell aimed to develop his questionnaires to measure these L factors
since by so doing these would also embrace all personality. The end result
of this work was the 16PF Test which measured the twelve L factors plus
four factors which regularly occurred among items Q 1 to Q4. This, grossly
simplified, is the basis for Cattell's claim that his factors embrace all
personality. He is certainly the first factor analyst to attempt to sample the
whole field of personality traits, although it must be pointed out that it is
likely that his initial reduction of the personality sphere, necessary given
the computational difficulties offactor analysis at that time, were too severe.
Reliability of the scales The reliability of the scales is far from satisfac-
tory. Ten of the sixteen scales have reliabilities lower than 0.7, which is,
as has been seen, the usually accepted minimum. Furthermore the parallel
forms of the test have low reliabilities, thus making it difficult to regard
them as equivalent.
Simple structure Cattell (e.g. 1978) has been responsible for many of
the arguments in favour of simple structure rotations and he and his
colleagues have developed many of the statistical methods involved.
Consequently the factor analytic procedures are technically extremely
good, although, as will be argued later in this chapter, other researchers
have found it difficult to replicate the factors.
Clinical evidence There is considerable clinical evidence, such as the

scores of different neurotic groups, in support of the construct validity of
the scales. This research can be found summarised in the handbook to the
test (Cattell Eber and Tatsuoko, 1970). This will be discussed in Chapter 9
of this book where the application of the psychometric view of personality
is scrutinised.
Occupational evidence There is also considerable support for the

validity of the test from occupational differences on the sixteen scales and
from correlations with occupational success. These factors are widely used
both in Great Britain and America in personnel selection (see Herriot,
1989). These results will again be discussed in some detail in Chapter 9.
Educational evidence From educational studies there is some support

for the validity of these Cattell factors, notably in the correlations with
academic success. Introverts tend to do better at higher education as do
mildly anxious students, presumably because they worry if they do not do
well.
Genetic facton There is a good body of evidence suggesting that these

personality factors are considerably determined by genetic factors. This
evidence has come from biometric studies of the personality factors and
Cattell (1982) has again been a pioneer in the complex mathematical
analyses demanded by biometric methods: work discussed in Chapter 7.
Construct validity It is obvious from the assembly of all these points

that there is a huge body of research evidence in favour of these Cattell
factors which, given their origins, would seem to support Cattell's claims
that these are the definitive factor analytic, psychometric description of
personality, and colleagues of Cattell at Illinois, and lately in Hawaii, do
strongly articulate this case, as exemplified by Cattell and Johnson (1986).

Nevertheless it has to be said that few outside his group are convinced.
There are many reasons for this divergence of views. Some are relatively
trivial. Thus it is clear that many psychologists who neglect the work are
not prepared to master the multivariate statistical methods which underpin
it and are thus unable to follow the arguments concerning rotational
procedures, just for example. Furthermore, Cattell invents neologisms to
describe his facton because he does not want their meaning to be confused
by the connotation of words in everyday use. These again create confusion
to those outside the system.
However, there are some more substantive reasons and these must now
be discussed. It has always been found difficult by researchen not associated
with Cattell to reproduce the Cattell facton. As was mentioned in the
previous discussion of the Guilford facton, Eysenck and Eysenck (1969)
failed to reproduce the Cattell set. However, Cattell (1973, 1978) put these
failures down to inadequacies in the factor analytic methods used; problems
which were discussed in the previous chapter. The present writer was
convinced by these arguments until together with Paul Barrett he under-
took a number of factor analytic studies of questionnaires including the
16PF and EPQ (Eysenck and Eysenck, 1975) tests; work reported in Kline
and Barrett (1983).
They found that simple structure analyses of the Cattell facton yielded
only seven facton and these were made up of items from a variety of the
Cattell facton. Furthermore many items loaded on more than one factor.
At the second-order two facton were clear-cut,- anxiety and extravenion.
They were forced to conclude that the sixteen facton postulated by Cattell
were not the most panimonious ~r elegant description of the penonality
sphere.
A few comments should be made about this study. We were careful to
ensure that it reached the technical standards advocated by Cattell. Thus
there was a high ratio of subjects to variables and the number of facton
was selected by the Scree test. The most efficient oblique rotations were
used, as found by Hakstian (1971), and the hyperplane count was tested
(all procedures discussed in Chapter 3). Indeed confirmatory analysis was
employed to attempt to find sixteen facton. Finally it should be pointed
out that the EPQ was subjected to these same techniques and three facton
were found as predicted; work discussed later in this chapter.
As has been indicated, the 16PF test embodying Cattell's facton
represents the results of the largest research programme into the psycho-
metrics of penonality which has ever been devised and it has been
conducted with almost unequalled technical brilliance. How is it, there-

fore, that it now appears that the factor structure of the test is wrong?
There are several reasons which might account for this and tht:,se must
be briefly discussed. First it could well be the case, as Howarth (1976) has
argued, that the twelve life factors based on the ratings were incorrectly
identified. This work was carried out before the introduction of computers
and utilised cluster analysis, a simplified form of factor analysis. Such an
error would be reflected in the 16PF test which was aimed at the L factors.
This argument is supponed by the fact that, as has been seen, modem
studies of ratings yield the big five factors (McCrae and Costa, 1987) and
more recent factorisations of the Cattell and other test items by Noller et
al. (1987), Boyle (1989) and Amelang and Borkenau (1982) have all
concluded that much of the reliable variance in the 16PF test is taken up
by these five factors. This work will be further discussed later in this chapter.
In addition, as was argued above, it is possible that the semantic
personality sphere was too constricted before ratings were begun.
In brief, therefore, it may well be that the failure to replicate the sixteen
factors in the Cattell system arises from the fact that the whole basis of the
analysis was wrong. Cenainly it is the case that if the original factor analyses
of the 16PF are examined (see Cattell, 1957), Cattell was prepared to accept
rather low loadings as evidence that the correct variables loaded the factors.
If, however, these arguments are accepted it is still necessary to explain
why it is so widely used in occupational selection. As has been fully
discussed in Kline (1992a), there are several reasons to account for this
anomaly. First, the two largest second-orders, extraversion and anxiety, are
present in the 16PF, as are shon scales of tough-mindedness (I) and
conventionality (G) which are also pan of the big five. Thus the influence
of these factors may account for the multiple correlations between the
factors and occupational success. Secondly, on sixteen scales it would be
surprising if there were not some differences between occupational groups.
Since there are few good theories of occupational choice it is not difficult
to develop reasonable convincing post hoc arguments for the results. The
third argument concerns the scientific naivete of most occupational selec-
tion processes. Applicants are selected and if found satisfactory it is assumed
that the selection methods were adequate. However, there is no follow-up
of those who were rejected. Indeed it may well be the case that in
management jobs, where all applicants are highly qualified and well-
experienced, it would matter little which were taken on the payroll.
Conclusions
From all these arguments it must. be concluded that the sixteen Cattell
factors do not represent the most simple and efficient factor analytic
description of personality. This is not to denigrate the contribution of
Cattell to this field. His contribution to the factor analysis of personality
showed how the subject should be tackled and his advances in methodo-
logy have enhanced the whole of psychometrics.
The Objective Analytic Test Battery (OAB) (CatteD and

Schuerger, 1976)
Before leaving the work of Cattell on temperament, a little more needs to
be said about the OAB, which was mentioned in Chapter 2 as being one
of the few published objective tests. Ishall not say much about it because
the study by Kline and Cooper (1984b) demonstrated that it was not a valid
test, at least in Great Britain. Nevertheless it deserves a brief description
here if only because of its ingenuity and originality.
It claims to measure ten source traits: self-assertion, independence,
evasiveness, exuberance, regression, anxiety, realism, self-assurance, exvia
(extraversion) and discouragement.
There is a huge number of objective subtests which are too numerous
to describe in detail here but include, for example: estimates of how good
performances are; rapid calculations ofarithmetic; picture memory; human
nature where subjects indicate their agreement with common beliefS about
people; how long does it take? - where subjects indicate how long they
would take to do certain things.
It is clear that this is an ingenious and truly objective test in that
deliberate faking is virtually impossible. However, because there is little
support for its validity, I shall say no more about the OAB (more details
may be found in Kline, 1992a) but pass on to the work ofEysenck.
The work of Eysenck

Eysenck, like Cattell, has made a huge contribution to the psychometric
study of personality. He has more than 1,000 publications to his credit but
the publications most relevant to this chapter are Eysenck (1967) and
Eysenck and Eysenck (1976), although other research will be cited as
necessary.
Eysenck, over the years, has produced a number of personality tests,
each an improvement on the one before. Thus there was the original
Maudsley Medical Questionnaire which measured neuroticism (N), the

Maudsley Personality Inventory measuring extraversion (E) and neuroti-
cism, as did the Eysenck Personality Inventory and the more recent
Eysenck Personality Questionnaire, the EPQ (Eysenck and Eysenck, 1975)
which was the defmitive measure of the Eysenck factors, extraversion,
neuroticism and psychoticism (P). However, in 1992 a yet newer version
has been produced, the EPQR (Eysenck et al., 1992). However, since
almost all the research into the nature of these factors has been conducted
with the EPQ I shall describe this test although the new aspects of the
EPQR will be discussed at the end of this section.
The original basis of the N factor was ratings of patients suffering from
neurotic disorders. These were factored and an N factor emerged which
was measured in the MMQ. The extraversion factor had its origins in the
study of the Guilford factors and the psychoticism factor had been observed
in studies of abnormal subjects, particularly psychotic patients and certain
criminal offenders, but was only successfully put into questionnaire form
in the EPQ. Descriptions of the basic research, culminating in these EPQ
factors, can be found in Eysenck (1967) and Eysenck and Eysenck (1976).
In this section I shall describe the EPQ because this is the measure of
the factors which has been used in attempts to validate them and to explore
their psychological significance. I shall then summarise this research into
the nature of the EPQ factors which distinguishes the work of Eysenck
from most factor analysts (except of course Cattell), who are content to
produce tests. It is this research which gives psychological meaning to the
EPQ variables and ensures that the factors are far more than sets of
homogeneous items.
The Eysenck PersoMiity Questionnaire

Four variables are measured, as set out below:
1 Extraversion-introversion. The extravert is cheerful, sociable, outward
looking while the introvert is withdrawn, quiet and inhibited.
2 Neuroticism. The high neurotic is worrying and anxious, the low scorer
the opposite of this.
3 Psychoticism. The high scorer is tough-minded, ruthless, likes powerful
sensations and lacks empathy. Criminals are high scorers and males are
higher than females. In the EPQ, normals scored very low on this factor
but this distribution has been improved on the recent EPQR, thus
allowing more precise factor analytic studies with normal populations.
4 Lie scale. This screens out those who give socially desirable responses.
Reliabilitie1 The internal consistency reliabilities of these scales are all

beyond 0. 7 and test-retest reliabilities are also beyond this figure except for
the P scale; this latter caused by the low scores of females on this variable.
For a penonality questionnaire these reliability coefficients are excellent.
Validity of the scales These are unquestionably the best validated

facton in the psychometrics of penonality. This arises from the extensive
experimental work of Eysenck and colleagues, as well as many other
psychologists, into the nature of these facton. This research will be
summarised under a number of headings.
1. Factor structure There can be no doubt that the factor structure of the
EPQ is exacdy as it should be with each of theN, P, E and L items loading
their respective scales. Thus Kline and Barrett (1983) demonstrated in a
simple structure oblique rotation of the intercorrelations between the
virtually perfect separation ofN, E and P items. Factor loadings were high
and the construct validity of N, E and P was supported most strongly.
Helmes (1989), a specialist in the factor structure of the EPQ, subjected
the items to a confirmatory analysis, the target matrix being derived from
the marking key. He confirmed the results of Kline and Barrett (1983),
although some ofthe P items were weak, because oftheir low endonement
rate, a failing which was mentioned above and which has been remedied
in theEPQR.
Thus there are three clear facton in the EPQ. Their psychological
nature has been determined by research described below.
2. Physiology Eysenck (1967) has described the physiological basis ofthese

facton. Extravenion has been related to the arousability of the central
nervous system, neuroticism to the lability of the autonomic nervous
system and psychoticism to androgen level. The extravert is lowly aroused.
This accounts for the fact that extraverts enjoy noisy situations such as pubs
and parties and are highly sociable. This is why extraverts are easily bored
and cannot stand monotony. Without noise the extravert is likely to nod
off. Introverts who like quiet punuits, such as reading and thinking, are
'always highly aroused and further environmental stimulation is simply too
much. In the occupational field, for example, it has been shown that
extraverts are poor at monotonous tasks, making erron and that introverts
should be selected for them. Many of the correlates of extravenion are to
be found in Eysenck (1970).
The autonomic nervous system controls involuntary responses such as
heart rate, sweating, digestion and the dissemination of hormones
throughout the body. Lability means that this system is easily activated and
then inhibited, what is called sympathetic and parasympathetic activity.
That this is involved in the neuroticism or anxiety factor is no surprise.
The rapid mood swings, the stomach contractions, pallor and sweating, all
associated with anxiety, make this hypothesis likely. Thus the individual
high on neuroticism has a highly labile system whereas the stolid, phleg-
matic low scorer has an autonomic system which does not much fluctuate.
Eysenck (1967) contains much of this physiological evidence.
The physiology of P has not been studied to the same extent as the
other variables and the implication ofhigh androgen level in high P scorers
is considered to be more of a tenable hypothesis than one confirmed by an
overwhelming body of evidence. However, the sex differences and the
fact, that many traditionally feminine characteristics, especially empathy
and tender-mindedness (see Guilford's masculinity factor), load this factor,
make this hypothesis not unlikely.
3. Heritability ofE, Nand P. Eaves et al. (1989) showed in a study ofSOO

twins using biometric analysis that these variables had a high heritability
index with about 70 per cent of the population variance being determined
by genetic factors. This and other similar work will be discussed in Chapter
7 on the heritability of personality where the significance of the findings
for personality theory will be examined. Here, however, it is sufficient to
point out that the fact that these factors are strongly genetically determined
means that it is unlikely that they are simply statistical artifacts of the factor
analysis. High heritability indices always confer psychological significance
on variables.
Here, it should be pointed out that in factor analysis a factor may always
be produced if sufficient similar items are included in a test. Cattell (1973)
refers to such factors as bloated specifics. This is why in our discussion of
factor analysis, it was argued that all emerging factors, even from simple
structure analyses, should be identified with reference to some external
criterion. Bloated specifics could never have high heritability indices.
4. External criteria The Eysenck factors are related to a wide variety of

external criteria in accord with their claimed psychological characteristics.
These findings are fully discussed in Chapter 8 on the application of
psychometric personality factors in clinical, educational and occupational
psychology. Here it is sufficient, as a demonstration of the construct validity
of the factors, to note some of the most important results, namely that these
factors are implicated in neurosis (Eysenck, 1961), smoking, criminality,
political allegiance and school learning (see Modgil and Modgil, 1986, for
interesting summary discussions of many of these and similar issues).
Conclusions concerning the Eysenckfactors

All these lines of evidence - the replicable clarity of the factor structure,
the correlations with external variables, the high heritability indices and
the dear implication ofphysiological structures- finnly support the validity
and psychological importance of these factors.
This conclusion receives further support from the survey of normal
factors in questionnaires by Kline and Barrett (1983) who argued that these
three factors and a fourth one of obsessionality were the only dear and
replicable factors to be found. More recent work on the big five personality
factors confirms this view, where these three factors are seen as three of
the big five (McCrae and john, in press); work to be discussed at the end
of this chapter.
The factors which have so far been discussed have emerged from the
pioneers in the field of the factor analysis of personality, whose factors have
been investigated, as has been seen, in respect of a wide variety of external
criteria. However, from the many other factor analysts, who have provided
their own sets of factors, I shall briefly describe the work ofComrey, who
attempted to improve the psychometric characteristics of personality
inventories and whose factors are highly reliable and stable, and work on
the big five factors, research which has been mentioned throughout this
chapter.
The Comrey factors

Comrey (1970) developed a new set offactors precisely because there was
so little agreement between the factors advocated by Cattell, Eysenck and
Guilford and because he considered that, as has been argued, there were
psychometric deficiencies in their scales. The factors in the Comrey scales
are:
Trust vs. Defensiveness
Social conformity vs. Rebelliousness
Emotional stability vs. Neuroticism
Empathy vs. Egocentrism
Orderliness vs. Lack of compulsion
Activity vs. Lack of energy
Masculinity vs. Femininity
Extraversion vs. Introversion

The reliability of the scales is high, all beyond 0.8, and the scales cannot
be faulted on this point.
Comments on the Comrey factors

One of the weaknesses of many scales, according to Comrey, resided in
the fact that the individual items were unreliable, being either trichoto-
mous or dichotomous. This was remedied in the Comrey scales by using
seven point rating scales for responses and by factoring the correlations
between subsets of similar items (factored homogeneous item dimensions,
FHIDs) rather than individual items. This certainly accounts for the high
reliabilities.
Examination of the scales suggests that the main factors are measured in
this test. Thus extraversion, neuroticism (or anxiety), tough-mindedness
and conformity can all be found. Clearly the variance cannot be much
different from that of the three scales which have been scrutinised in this
chapter.
Cattell (1978) has criticised the rotational methods used by Comrey in
this test as being unlikely to reach simple structure and it is interesting to
examine the joint analyses of this test with the other important personality
factors. Thus Cattell (1973), citing research by Barton, claimed that the
Comrey factors were not simple structure factors but that there was good
agreement between them and the Cattell second-orders. Noller et al.
(1987) factored the Comrey, Cattell and Eysenck factors, using Comrey's
unusual rotational procedures, and found only seven factors, five of which
appeared to be the big five factors which have been mentioned throughout
this chapter. Boyle (1989) essentially confirmed the big five factors with
the same data but using more orthodox methods.
Conclusions
These Comrey factors are stable and reliable but are almost certainly not
simple structure factors. The variance in the test is highly similar to that of
the other personality questionnaires but at best, re-rotated, it measures the
largest second-order factors in the personality test realm. As they stand the
Comrey factors are not the simplest and most elegant personality factors.
The big five facton

From this discussion of the most important psychometric factor analytic
studies ofpenonality, it has become evident that what'are usually referred
to as the big five facton have been claimed by various researchen to
underlie the variance in all these tests. In other words the best factor analytic
description of penonality is that there are five penonality facton and that
most tests even when they were not designed specifically to measure these
facton in fact do so. To conclude this chapter, therefore, I shall examine
the evidence for postulating the big five.
Description of the big five factors

1 Extravenion.
2 Agreeableness.
3 Conscientiousness.
4 Neuroticism.
5 Openness to experience.
These are the names given to these factors by McCrae and John (in press)
in their most recent account of the system, although, as has been seen,
other names are sometimes used. For example, agreeableness is similar to
tender- vs. tough- mindedness and openness to experience equates with
conventionality. The precise names are not important compared with the
identity of the factors in the various studies. As these authors argue, these
factors have been found in self-reports, ratings, natural languages, theore-
tically-based questionnaires, English, Dutch, German and Japanese adult
samples Qohn, 1990). In addition they appear to endure across decades in
adults (McCrae and Costa, 1990).
Tupes and Christal (1961) noted that five factors seemed to emerge
from studies of personality in which ratings were involved; a claim that
was supported by Norman (1963). Goldberg (1983) also showed that
analysis of descriptive terms in natural language seemed to reveal about five
factors. This analysis of natural language was, as has been seen, the original
basis of Cattell's 16PF test. Furthermore, analysis of the leading personality
tests also seemed to reveal that about five factors accounted for the reliable
variance. McCrae and Costa in various studies have been responsible for
much of this work. Thus the big five were to be found in the Eysenck
scales (McCrae and Costa, 1985), the Guilford scales (Amelang and
Borkenau, 1982), the Comrey ,and Cattell scales (N oiler et al., 1987), the
Myers-Briggs Type Indicator (McCrae and Costa, 1989a), the MMPI
(Costa et al., 1986), the Personality Research Form (Costa and McCrae,
1988), the last three tests being described in Chapter 5, and even in a test
of vocational interests, the Holland Vocational Preference Inventory
(Holland, 1985).
Such a consensus, from ratings, factor analytic and other personality tests
has led not only the main proponents of the big five, Costa and McCrae,
to claim that these factors underlie the factors among personality traits, but
other researchers in this field have accepted this view (e.g. Digman, 1990).
Where there is agreement among different tests and with different statistical
methods, it seems difficult to reject the claims. Imperfect rotational
procedures, for example, add in error and are most unlikely to result in
such striking agreement. Certainly it is not difficult with questionnaire
items to find a five-factor structure as Kline and Lapham (1991) showed
in the development ofthe PPQ, a personality test for occupational selection
and designed to measure these big five factors.
NEO Personality Inventory (Costa and McCrae, in press)

Devised by McCrae and Costa, this NEO ought to be the supreme
personality inventory. Its rationale springs from the findings that five factors
are common to ratings, natural language and inventories and it seems as
our references demonstrate to measure the big five factors. Certainly these
test constructors should be measuring the right variables.
Reliability of the scales

This is sufficiently high, all beyond 0.7, to make the test psychometrically
efficient.
Evidence for validity

The evidence for the validity of these scales has been discussed. It resides
in the factor analytic studies with other tests. However, as has been argued
by Kline (1992a), this evidence is not quite as strong as it first appears. Thus
it could be the case that these factors in the NEO did account for the
variance in other tests but that they were not the factors claimed by the
authors.
This need for external validation of the factors was answered to some
extent by an ingenious factor analytic study in which in a sample of nearly
1,000 subjects the NEO factors were located in factor space relative to a
form of the NEO completed by peers and spouses (observations by others
on the NEO items), adjective check lists and biographical facton (McCrae
and Costa, 1989b).
However, a form ofProcrustes rotation was used which has been shown
to be able to hit target matrices in random data and correlations with other
similar measures are not as convincing as are correlations with criteria
beyond the domain of tests.
Conclusions
There is little doubt that the NEO inventory is about the best measure of
the big five facton, which are clearly implicated in the factor analytic
penonality tests which have been described in this chapter. Similar facton
appear in ratings and natural languages. Nevertheless, the validity of these
NEO scales is not supported by correlations with external criteria, as is the
case with the EPQ measures of these facton. This, of coune, is a much
newer test and it is to be hoped that such research will be punued.
Thus from this survey of the factor analytic study of penonality there is a
powerful case for three facton: extravenion, anxiety or neuroticism and
tough mindedness, all with convincing external validity. The two other
facton, openness to experience and conscientiousness, are well supported
in the realm of questionnaires, although their independence (they may be
aspects of obsessionality, see Kline and Barrett, 1983) needs further
research, as do their external correlates.
In Chapter 5 I shall consider the findings from penonality tests con-
structed by methods other than factor analysis, since their findings must
also form part of the psychometric view of penonality.
Chapter 5
Findings from other types of

psychometric tests
Although factor analysis has been the statistical method which has led to
the establishment of clear factors in the field of personality, as was
demonstrated in the previous chapter, nevertheless, as was pointed out in
Chapter 2, tests can be developed by other methods, not involving factors.
In this present chapter the substantive findings from the best tests of this
kind, all of course aspects of the psychometric view of personality, will be
considered.
Minnesota Multiphasic Personality Inventory, the MMPI

(Hathaway and McKinley, 1951)
I shall begin with a study of the research with the personality questionnaire,
which is undoubtedly the most widely used and researched of all inven-
tories, there being more than 10,000 published articles, books and chapters
about this test, let alone unpublished doctoral dissertations (Eysenck, 1989).
Recently, however, there has been a substantial revision of this test- the
MMPI-2 (Graham, 1990)- and this will also be discussed. However, since
this is so recent, most of the discussion will refer to the older test except
where specifically stated.
Basic information about the test

-The scales were constructed by the criterion keyed-method, i.e.
items were selected if they could discriminate clinical groups, a
method of test construction discussed in Chapter 2.
-Nine clinical variables are measured by the basic MMPI although
more than 200 other scales have been developed (Dahlstrom and
Walsh, 1960). These are:
68 Other types of psychometric tests
Hs. Hypochondriasis; D. Depression; Hy. Hysteria;

Pd. Psychopathic deviate; Mf. Masculinity/Femininity;
Pa. Paranoia; Pt. Psychasthenia; Sc. Schizophrenia; Ma. Hypomania
Si. Social introvenion.
-The test contains 566 items of the 'True' /'False' variety. However,
various short venions of the test have been developed.
- Reliabilities. The internal consistency reliabilities of these scales are
far too low for psychometric efficiency. Certainly they are not
homogeneous, as Graham (1990) makes clear.
Discussion of the test

Clearly 10,000 published studies could not be fully examined even in a
large book. However, I shall concentrate upon those researches critical to
the thesis of this chapter, namely those which bear on the substantive
findings from the test. Before this can be done, however, it is necessary to
discuss a number of difficulties with this test, problems which render many
of the findings somewhat dubious. In fact, many of these have been dealt
with elsewhere in this book (Chapter 2) and thus they can be briefly listed
here.
The fint problem concerns the criterion-keyed method of test con-
struction, which, as has been argued, leads to scales of little psychological
meaning since the criterion groups may differ on a number of variables.
This makes the psychological interpretation of research findings difficult.
This method also allows the same items to appear in different scales. When
this occun, correlational and factor analyses are difficult to interpret for
obvious reasons.
The low reliability of the scales was mentioned, a phenomenon which
is always likely with criterion'-keyed tests and which highlights again the
lack ofhomogeneity and thus meaning of the scales.
There is a further difficulty with the MMPI which stems from the fact
that it was developed as a clinical instrument for use with abnormal subjects.
While the test discriminates among abnormal groups, mainly because many
of the items refer to neurotic symptoms, when it is administered to normal
subjects there is far less variance and scores on the scales tend to be low.
Most normal individuals, by definition, have less symptoms than those
receiving psychiatric treatment. As was made clear from Chapter 3 on
factor analysis, a lack of variance affects the clarity of the emerging facton.
These problems are not such that it is impossible to investigate the
Other types of psychometric tests 69
psychological meaning of the MMPI but they do make research more

difficult and the results have to be treated with some caution.
The simplest approach to uncovering the psychological meaning of the
MMPI is to factor analyse the items (rather than the scales which are
probably not unifactorial) in the test. With 566 items more than 1,000
subjects would be required for a sound factor analysis, from the viewpoint
of subject numbers, as was discussed in Chapter 3 when the criteria ofgood
analyses were set out. This is important because many of the factor analytic
studies of the MMPI items fail to achieve simple structure, thus making
their results of dubious worth.
However, the investigation by Johnson et al. (1984) should be men-
tioned since it involved 11 ,000 subjects, thus reducing statistical error to a
minimum. Twenty-one factors emerged which were named from the
items loading on them, although it must be noted that there was no external
validity for their identification. The factors included: anxiety, psychotic-
ism, extraversion, paranoia, psychopathic deviation and psychasthenia.
These six factors have been selected because the first three factors have the
same names as three of the big five factors discussed in the previous chapter,
although it must be stressed that there is no empirical evidence that they
are those factors. The last three are similar to the three of the abnormal
Cattell factors found by Krug (1980), mentioned in the previous chapter
and forming a part of the CAQ. Despite the large numbers of subjects and
the fact that there appears to be some similarity with well-established
factors, this study by Johnson et al. (1984) does not provide a definitive
factor structure of these MMPI items however, partly because there is no
external validation of the factors.
Costa et al. (1985, 1986) carried out a factor analysis of the MMPI items
in a large sample of non-psychiatric patients and attempted to relate the
emerging factors to other scales, thus validating, to some extent, the factor
labels. Nine factors emerged: neuroticism, psychoticism, masculinity,
extraversion, religious orthodoxy, somatic complaints, inadequacy, cyn-
icism and intellectual interests. These factors were correlated with the big
five and Costa et al. (1986) argued that these were to be found in the MMPI
variance. Two points should be noted about this conclusion. If it is the
case that the big five factors are to be found in the MMPI items, they can
certainly be measured more efficiently by other tests. However, there is
some other variance in the MMPI, probably related to abnormal person-
ality, and this needs further clarification.
In the CAQ (Krug, 1980) the factors found by Cattell and Bolton (1969)
in their study of the MMPI items were incorporated into the test. These
were: paranoia, psychopathic deviation, schizophrenia, psychological
inadequacy and psychasthenia. These however, although clearly marked

as facton, still require external validation before they could be regarded as
substantive facton of abnormal penonality. Nevertheless they demand
research.
Conclusions .from factor analyses of MMPI items

The facton derived from the MMPI items are of two kinds, normal and
abnormal, as might be expected from a penonality test derived from
discriminations among clinical groups. The normal facton are probably
similar to the big five while the abnormal facton require external valida-
tion, although it should be said that the paranoia, psychasthenia and
psychopathic deviation facton seem to replicate well and make good
psychological sense.
Factoring the MMPI scales

It might appear more simple to factor the MMPI scales rather than the
items and indeed this has frequendy been done. However, the difficulty
of item overlap, the low reliability of the scales and even selecting which
scales to choose of the 200 which have been developed, makes the task
somewhat problematic. However, despite these difficulties there is consid-
erable agreement as to the facton in the MMPI scales, as Friedman et al.
(1989) point out. There are two replicated facton- anxiety and repression.
The fint of these is not unexpected.
Conclusions
Despite the huge amount of research conducted with the MMPI, it has to
be concluded that it has added litde to our knowledge ofpenonality. This
is because it was developed by criterion analysis of an item pool before
factor analysis was generally available. A heroic research effort has resulted
in a test which may be useful for screening, but its psychometric inade-
quacies, its low reliability and lack of a clear factor structure mean that it
should after more than half a century be gracefully retired. It is a test
developed by outmoded technology.
However, these problems were not unknown to some usen of the
MMPI and, as was stated above, a new form of the test, the MMPI-2, was
developed and this will now be discussed.
MMPI-2 (Butcher, 1990)

This version of the test was developed, according to Graham (1990), to
improve the samples from which the original clinical scales were derived,
to modify the items some of which were now obsolete in terms of content
and to widen the behaviour to which the items referred (e.g. the use of
drugs) .
It must be noted that these aims do not overcome the deficiencies of
the test from the viewpoint of contributing to a knowledge of psychology.
Indeed the fact that it was intended that the MMPI-2 resembled the old
test as much as possible, but in an improved form, suggests that the lack of
a clear factor structure will still reduce the psychological meaning of the
scales. In fact, there is still item overlap and although the scales are more
reliable than the originals they are far from homogeneous. Although no
factor analyses have been carried out, Graham (1990) argues that anxiety
and repression are likely to be the main factors in the scales.
In brief it must be concluded that the same objections to the MMPI
apply to MMPI-2 and that this test cannot contribute much to the
psychometric view of personality.
Tests based upon theories and variables derived from tests

In this section of the chapter 1shall consider the findings from two different
kinds of personality inventories. First 1 shall scrutinise two well-known
tests which are based upon personality theories, those ofJung and Murray.
Then 1shall consider the evidence for two apparently important personality
variables, which have their origins in personality testing- the authoritarian
personality and locus of control.
Most theories of personality, as was made clear in the opening chapter,
lack rigorous, scientific evidence in their favour. Many of the most famous
theories are clinical, such as the various versions of psychoanalysis, and
although appealing, to endorse them as they stand, is strictly irrational. For
this reason, tests which measure the variables of such theories are particu-
larly important in the scientific study of personality since if they are
validated, this ipso facto supports the theory. Thus the scrutiny ofthe validity
of such tests scrutinises these theories.
The Myers-Briggs Type Indicator (Briggs and Myers, 1962)

This test claims to classify individuals according to the Jungian theory of
eight personality types and an extended typology (Jung, 1949):
1 Extraverted thinking 2 Introverted thinking

3 Extraverted feeling 4 Introverted feeling
5 Extraverted sensing 6 Introverted sensing
7 Extraverted intuition 8 Introverted intuition
The scales are reasonably reliable and thus all turns upon the validity of the
test.
Test validity Some continuous scores are obtainable from this test but
these are not relevant to this chapter and I shall ignore the studies of their
validity. What is critical is whether this test can classify individuals into
meaningful groups which resemble those described by Jung. However, it
should be noted that McCrae and Costa (1989) have argued that in fact
even in this test much of the variance is accounted for by the big five. This,
of coune, runs counter to the validity of the test as one ofJungian types,
although correlational studies are difficult since some of the scores in the
MBTI are ipsative.
Stricker and Ross (1964) examined the distribution of the MBTI scores
but found no evidence for typologies of any sort - bimodal rather than
continuous distributions of scores would be required. In fact, with multi-
detennined variables typological rather than continuous distribution would
be highly unlikely. In nature typologies are the exception.
Broadway (1964), in a paper which can be found in Vetter and Smith
(1971), penuaded twenty-eight Jungian analysts to classify themselves into
types and take the MBTI. There was full agreement for introvenion and
extravenion and better than chance classification for sensation and intui-
tion. It should be noted that this classification was not exactly into the
categories above but is acceptable Jungian theory. This appean impressive,
especially in respect of introvenion--extravenion, but it is by no means
conclusive support.
Thus it is quite possible to classify oneself correctly as an introvert or a
sensationist without implying typologies of any sort. The classification
could simply represent one's position above or below the mean. This is
highly important in the case of extravenion since, as has been shown, there
is general agreement that there is such a factor. The Jungian concept is
quite different involving a typology. Furthermore, we do not know the
quality of these Jungian analysts or the extent to which they are a
representative sample. Thus these findings are not convincing.
The Dynamic Personality Inventory (Grygier and Grygier, 197 6)

A brief mention should be made of this test which attempts to measure
Freudian personality variables, e.g. oral and anal characteristics, as described
by Freud in his general psychosexual theory.
There are thirty-three Freudian variables which are measured with
varying degrees of reliability and which, if shown to be valid, would
provide psychometric support for Freudian personality theory. I have
reviewed the validity of this test in great detail in Kline (1981, 1992a) and
it is sufficient to summarise the results here.
Most of the factor analyses of the DPI cited in the handbook to the test
have to be treated with caution because they fall short of the criteria for
satisfactory factor analyses. What is required is the location of these
Freudian scales in filctor space in order to investigate their construct
validity. Kline and Storey (1978) attempted this rotating to simple structure
the DPI scales together with the Cattell and Eysenck factors (see Chapter
4) and measures of authoritarian personality which are described in a later
section of this chapter.
From the viewpoint of this chapter, the interesting finding was that a
factor emerged which loaded on the measures of obsessional traits and
authoritarian personality, a factor which Kline and Barrett (1983) showed
was highly important in questionnaires and which is similar to the conser-
vatism and conscientiousness factors of the big five . It was also clear that
the DPI measured quite different factors from those ofEysenck and Cattell,
although there was little evidence to support their validity as measures of
psychosexual variables. It cannot be said, in brief, that this test provides
psychometric confirmation of freudian theory, or that the factors contrib-
ute to the psychometric view of personality.
Jackson Personality Research Form, the PRF ljackson, 1974)

This test is based upon Murray's (1938) theory of needs and presses which
postulates that a large number of needs (of which the PRF measures the
twenty most important) account for human behaviour. Thus if this test
were shown to be valid, these needs could be said to be part of the
There are various forms of the PRF but I shall discuss the results with
Form E which contains all the scales developed by Jackson in the shortest
form, and with a vocabulary level suited for general adult use.
The twenty needs of the PRF are: abasement, achievement, affiliation,
aggression, autonomy, change, cognitive structure (need for precision),
defendence, dominance, endurance, exhibition, harm-avoidance, impul-

sivity, nurturmce, order, play, sentience (need for physical sensations),
social recognition, succorance (need for support), understanding. In addi-
tion, there are two scales to assess social desirability and careless responding.
Test construction It is generally agreed that the PRF is psycho-

metrically one of the best tests. Given the relatively short scales of Form
E, the reliability is high and the item analyses were technically highly
sophisticated, as is fully discussed in Kline (1992a), although it is difficult
to see why factor analysis was not used. Thus all turns on the validity of
these scales. In the manual to the test there is no evidence presented for
validity other than correlations with ratings for the other forms of the PRF,
which is litde more than face validity. Other studies are required.
Nesselroade and Baltes (1975) factored the scales (not the items) in the
PRF and the HSPQ (the adolescent version of the Cattell16PF, discussed
in Chapter 4) in a well-conducted research. Eight factors were found,
which fact alone fails to support the validity of these PRF needs. Since, as
has been argued, the validity of the Cattell factors is itself in doubt, there
is no need here to attempt to interpret these factors. In brief, this study did
not support the validity of the PRF.
Guthrie et al. (1981) factored the PRF in a sample of filipino students
and found six factors, which he claimed were common to American and
French samples. Again, this study does not confirm the validity of these
scales as measures of twenty independent needs. Finally, as has been noted,
Costa and McCrae (1988) found the big five factors in this test, a claim
supported by Digman (1990).
Conclusiow concerning the PRF Despite the technical expertise in

the item analytic construction of the PRF, it has to be said that there is no
evidence, as yet, that it is a valid measure of the twenty needs it purports
to measure. There appear to be about eight factors, at the second-order,
in this test and these are highly similar to the big five factors and the Cattell
second-orders. Thus the PRF is essentially measuring what the majority
of personality scales measure inter alia - the big five factors.
From this it is clear that this test cannot be used to support the Murray
system of needs and presses. Furthermore it appears to possess no clear-cut
factors of its own which have been externally validated or identified. Thus
its positive contribution to the psychometric view of personality is not
large, although given its technical expertiJe the fact that it fails to support
the Murray needs is useful negative evidence.
Thus the two most widely used tests, deliberately constructed on the
basis of theories ofpenonality, the MBTI and the PRF, are probably not
valid and psychometric support for these theories cannot be derived from
these tests.
The authoritarian personality

Adorno et al. (1950) carried out one of the most famous studies of
penonality in which a battery of tests ofvarious types were given to fascistic
subjects in an effort to discover the psychological basis of fascism and
anti-semitism, which at that time, for obvious reasons, were of particular
relevance and interest. Although, as Brown (1965) has pointed out, there
were considerable methodological problems in this work, the concept of
authoritarian penonality has lived on and it seems to have considerable
explanatory power. The authoritarian, sycophantic to her superion, ruth-
less to inferion, bound by rule and status, only doing her duty, is all too
recognisable, in hierarchical organisations. It is they who represent the
banality of evil, as Eichmann has been described.
The question which I shall briefly consider in this section, although it
is an enormous subject (see, for example, Stone and Lederer, 1991), is
whether in the light of modern psychometrics the notion of authoritarian
penonality is any longer viable. If it is, of coune, it is reasonable to regard
it as a major contribution from psychometrics and an important aspect of
the psychometric view of penonality.
As was mentioned above, the original study of the authoritarian per-
sonality was criticised on methodological grounds: the F scale which
measured authoritarianism was not balanced for yes and no responses, thus
allowing the influence of acquiescence to affect its scores; the F scale items
were biased by social desirability; the scales measured only right wing
authoritarian attitudes; the scores reflected the attitudes of the ignorant
ill-educated rather than the specifically fascistic. In response to these
criticisms there have been many attempts to produce new authoritarian
measures, free of these defects.
Kline and Cooper (1984a) .factored a number of authoritarian scales in
an attempt to locate the authoritarian personality in factor space. This, it
was hoped, would demonstrate whether it was a useful concept or whether
the variance was better accounted for by other factors. They showed that
there was an authoritarian factor which loaded on measures of obsessional
personality and on a measure of anal character. They concluded that the
authoritarian personality, was the social, attitudinal emanation of the
obsessional personality, the factor which Kline and Barrett (1983) had
shown was a large second-order factor among personality questionnaires.
It should be noted, in respect ofthis analysis ofthe authoritarian personality,

that this obsessional factor is highly similar to the open-mindedness and
conventionality of the big five.
Christie (1991) has an excellent review of most of the measures of
authoritarian personality and argues that one scale is superior in terms of
psychometric qualities and content to other tests of authoritarian person-
ality. This is the Right Wing Authoritarian Scale, the R WA, by Altmeyer
(1981), a highly reliable balanced scale, with good evidence of validity.
This test has twenty-four items to which subjects respond on a six-point
scale- disagree strongly to agree strongly.
Before drawing any conclusions about the authoritarian personality,
mention should be made of a highly similar concept - dogmatism
(Rokeach, 1960). Rokeach had strongly contended that the authoritarian
personality was a right wing form of a more general trait - dogmatism,
characterised by rigidity and a mind closed to new ideas which could be
found along the whole political spectrum. Rokeach, therefore, developed
a forty-item scale to measure rigidity of belief, of high internal consistency
but still of unproven validity. Certainly Kline and Cooper (1984a) found
that a counterbalanced dogmatism scale (Ray, 1970) correlated only low
with a counterbalanced F scale (Kohn, 1972), the scale which loaded
highest on the authoritarian factor. The work reported in Rokeach (1960)
indicates that dogmatism is an interesting variable. However, the factor
analytic research demonstrates that it is not identical with the F scale. It
would be useful if the dogmatism scale were rotated to simple structure
together with the main personality tests to identify its psychological nature.
However, it does not seem to be as important or robust a variable as the
authoritarian personality.
Conclusions concerning the authoritarian penonality The orig-

inal study of the authoritarian personality by Adorno et al. (1950) was a
highly impressive and convincing account of authoritarianism despite the
fact that it had psychometric methodological problems. Modern studies of
authoritarianism with improved measures support the concept of authori-
tarian personality. It seems to be the political or attitudinal aspect of the
obsessional personality, a syndrome ofpersonality traits typified by the need
for self-control, control of others, rigidity and conservatism. This person-
ality concept has been defined and refined by psychometric testing and it
must be considered a major contribution from psychometrics to an
understanding of personality.
Other types of psychometric tests n
Locus of control
Locus of control is variable which has no immediate referent in everyday
language, although there is a large number of psychometric tests which
purport to measure it. Thus if any of these were shown to be valid, the
concept of locus of control would be a contribution of psychometrics to
personality and, as was the case with authoritarian personality, an aspect of
the psychometric view of personality.
Locus ofcontrol is a concept which originated in social learning theory,
as it was developed by Rotter (1966). External locus of control refers to
the belief that the external environment rather than personal effort deter-
mines what happens. Internal locus of control is the opposite, namely that
outcomes are contingent on actions. Locus of control is held to be a
personality characteristic or trait which is broadly generalisable across a
variety of situations.
Nevertheless, as was pointed out in Kline (1992a), there is a conceptual
difficulty with locus of control which renders it of dubious worth although
it is widely used in social psychology. Thus Lefcourt (1991), in a chapter
which discusses many of the tests of locus of control, argues that measures
oflocw ofcontrol should be tailored to particular populations and concerns
and that these specific measures are more efficient than a broad measure of
the variable, a viewpoint which the major workers in this field all support
(e.g. Phares, 1976; Rotter, 1975). However, if this is the case, it suggests
that locus of control is not a broad trait at all, but rather a specific and
particular belief and is thus oflittle psychological interest. An example will
clarifY this point. The general locus of control dimension reflects a belief
that outcomes are detennined by external influences at one pole contrasted
with personal effort at the other. Thus such a scale should predict such
beliefs in fields such as health or success at work. However, in fact this is
not so. Specific measures relating to health and relating to work are
required. This suggests that there is no general factor and that the variable
is relatively trivial psychologically.
It is therefore of considerable interest, given this theoretical argument,
to examine the factor structure and the location in factor space of the
general locus of control variable, the specific tests being oflittle theoretical
or psychological interest.
Lefcourt (1991) in his lengthy review of tests oflocus of control claims
that the original locus of control scale (Rotter, 1966) is still the most used
and widely cited test so that I shall restrict myself to examining the
psychological meaning oflocus of control, as thus measured. Actually many
other more recent locus of control tests use items from this scale together
with modified and similar items so that the fact there are correlations
between the various measures is not really evidence for validity.
The Rotter 1-E scale This scale has twenty-three items and six filler
items. This suggests that, as was made clear in Chapter 2, the scale should
be reliable and this is so. Thus all depends on the validity of the test,
specifically its factor structure and its relationship to other well-known
factors.
It is instructive to examine the original item analysis of this scale. The
correlations of items with the total score are unusually small. Seventeen
items have correlations of0.29 or below, five are above 0.3 and only one
is above 0.4. This does not suggest that there is a common factor accounting
for the variance in these items.
Ashkanasy (1985) has reviewed much of the factor analytic work on the
items in this scale and the results can be easily summarised. There are
generally two factors with a large number of uninterpretable small factors.
The first factor loads on items phrased in the first person, the second on
items relevant to political and social institutions. All this strongly implies
that there is no factor oflocus of control and that the concept has arisen
through psychometric ignorance - collecting a set of face valid items and
calling it a test.
This negative evidence concerning the emergence of a locus of control
factor arises from the study of the items themselves. It is, however,
confirmed by research on the correlations and factorisations of the locus
scores with other tests. Thus Lefcourt (1991) admits that the scale correlates
positively with social desirability, a finding which creates little confidence
in the concept oflocus of control as an important personality variable.
A further study essentially demonstrates that there is little useful variance
in the locus of control scale. Thus Layton (1985) correlated the test with
the EPQ (see Chapter 4). He found that there were significant positive
correlations with N, P and L. Given the lack of a clear factor within the
items of the 1-E scale, it appears that two of the important Eysenck factors
and social desirability are all that is reliably measured by the test.
Conclusions It is clear that there is no common factor of locus of

control. Whatever is measured by these items is largely composed of
variance in other tests, especially neuroticism, psychoticism and social
desirability. Since, in addition, it was argued by Lefcourt (1991) that the
general locus of control factor cannot predict behaviour but that specific
locus of control scales are better, there seems no reason, theoretical or
psychometric, to posit a locus of control variable.
That this scale and other similar scales are widely used in social
psychology demonstrates the chasm between different areas ofpsychology.
It also demonstrates a collective madness that has overtaken various
branches of psychology in which studies exist in a private world, in this
case oflocus ofcontrol (see Kline, 1988, for a full discussion of this problem
in many different fields of psychology). Scales are used which correlate
with each other mainly because their items are highly similar if not
identical, and these correlations are cited as evidence of meaningfulness.
Such work is hermeneutical and of no scientific value.
In brief, locus of control is not a construct that is supported by
psychometrics. Indeed the contribution of psychometrics to this field is to
demonstrate the inanity of the concept, a contribution, however, to which
no attention has been paid.

In this chapter I have scrutinised the findings from psychometric tests, of
high repute and wide usage, which were constructed by methods other
than factor analysis. The MMPI, the personality questionnaire, with the
most citations in books and papers, and originally devised by criterion-
keying items in the discrimination of abnormal groups, was shown to be
disappointing in its substantive contribution to psychological knowledge.
This monument to empiricism was useful only for screening out abnormal
subjects and its recent revision, the MMPI-2, seems litde improvement in
this respect.
Two tests, the PRF and the MBTI, devised to measure variables in the
Murray and Jungian systems respectively, were shown to be of unproven
validity and, in all probability, measured litde variance beyond the big five
factors usually found in questionnaires. It was also shown that an interesting
but litde known questionnaire, the DPI, claimed to be measuring Freudian
psychosexual variables, had litde support for its validity although its
variance did appear to be different from that in most personality question-
naires.
Finally two variables, authoritarianism and locus of control, both
essentially products of psychometric testing, were scrutinised. The latter
was shown to have little support conceptually or empirically and it is a
concept which should be abandoned. Authoritarianism was shown, on the
other hand, to be a well-founded variable and to be the social, political
emanation of the obsessional personality, a personality syndrome which
resembles two factors in the big five, open-mindedness and conservatism.
From this study of empirical and theoretically derived tests which were
not based upon factor analysis, the only concept which has been sup-
ported is that of the authoritarian penonality. This is a variable or syndrome
of penonality traits which must form part of the psychometric view
of penonality and whose theoretical implications will be discussed in
Chapter 9.
Chapter 6
Personality dynamics
The psychometric view
In our opening chapter a distinction was drawn, it will be remembered,

between temperamental and dynamic personality traits. The former relate
to the manner of behaviour, the latter to determinants. Extraversion for
example, a temperamental trait, is reflected strongly in social behaviour.
As has been shown, the extravert is noisy, sociable, pushy, hungry for
stimulation. This should be contrasted with dynamic traits which motivate
or determine what we do. Thus, to take the same example, a person may
seek out a party because she feels lonely, the dynamic trait here being the
need for company or love.
As was also pointed out in the first chapter, the nature of human
dynamics is a matter of considerable dispute among personality theorists,
ranging from two drives, at one extreme - the Eros and Thanatos of
psychoanalysis- to twenty-three or more in the theories of Murray (1938)
and McDougall (1932), and even including the view of Skinnerians
(Skinner, 1953) that behaviour is best regarded as controlled by reinforcers,
although certain basic biological drives, such as hunger or thirst, are
acceptable to them.
This theoretical confusion concerning personality dynamics suggests
that it is an ideal field for the factor analytic psychometric approach to pick
out the most important factors. However, there are considerable difficulties
involved in the study of psychodynamics which must be discussed and
examined before the results of the psychometric investigation of this field
can be appreciated.
Distinction between temperamental and dynamic traits

There can be no doubt that there is a distinction between temperamental
and dynamic traits. Hunger acts as a drive to seek out food and nobody
would think of hunger as a personality temperamental trait. A person may
82 Personality dynamics
be described as hungry for power or f.une but again this is seen as a dynamic
trait, detennining behaviour.
In contrast to this, a temperamental trait such as obsessionality accounts
for how people do things- neatly, precisely, in a certain order and without
error. Similarly the anxious person checks train times, arrives early to
ensure there is no mistake and probably books a seat. This latter example
demonstrates that the distinction between temperament and dynamics is
blurred. It would be reasonable to say, in this instance, that she checked
her ticket because she was anxious. Thus clear temperamental traits can
have a dynamic aspect.
It might be argued that anxiety is a special case since, as is discussed
below, anxiety is not only a trait but is also a state (e.g. before an
examination) and states or moods definitely determine behaviour. Indeed
such an argument might even apply to the train whose imminent departure
had produced state anxiety. This, however, will not do, if we consider the
further distinction between states and traits.
Distinction between states and traits

Traits are stable characteristics unchanging over time. States are transient
and may last for a very short period. Anger and fear exemplify this latter
case although fear can be long lasting (timor mortis perturbat me) in which
case it becomes a trait- timidity. The timid person varies in fearfulness
(state fear increasing it from time to time) . Thus, although this distinction
between stable and transient, to discriminate states, is not absolute, in
general it makes good sense. Most states or moods, for the difference is one
of English usage rather than reflecting a psychological phenomenon, are
transient. Traits are always stable or long term. It makes no sense to talk of
a transient trait.
Thus to return to the distinction between dynamic and temperamental
traits it has to be admitted that it is not absolute although in the majority
of cases it is meaningful. Even the case of obsessionality could be seen as
causing the checking and listing typical of the syndrome, although this
seems to be a different sense of, or level of, causation from drives such as
hunger or thirst. Despite these problems, however, it is clear that there is
a useful distinction between dynamics and temperament.
It should be noted that one of the tests of temperament that was
discussed in the last chapter, the PRF, measured a set of needs. This test
could be conceived ofas a dynamic measure since needs by definition drive
behaviour. However, this test is usually regarded as a personality rather
than a motivational test and, as was seen, its variance considerably over-
Personality dynamics 83
lapped two of the best known sets of temperamental factors - those of

Cattell and the big five.
This introduction discusses some of the problems besetting research into
the dynamics of personality. These can be briefly summarised. First there
is considerable theoretical confusion concerning what are the important
dynamic variables, thus making it difficult to know what to test. Secondly
the distinction between dynamic and temperamental traits is not as sharp
as it first appears and a further distinction between states or moods and traits
has to be taken into account although even this is to some extent relative.
Factor analysis of dynamic traits

In the field of personality traits, as has been seen, the approach pioneered
by Cattell has been to sample the universe of temperamental variables and
to discover the most important factors by simple structure rotation. This
had been previously successful in the sphere of abilities (e.g. Cattell, 1971)
and this is the obvious method for clarifying human motivation.
From the introductory discussion of personality dynamics it is clear that
there are two categories of motivational trait: states or moods and drives.
These will be examined separately.
Factor analysis of states and moods

In addition to the general problems of research into personality dynamics
which have been discussed there are some more specific, practical difficul-
ties which create further problems for factor analysis.
States change over time. As has been argued, states are distinguished
from traits by their transience. This means, as Cattell (1973) has pointed
out, that any factor analysis which will reliably reveal states, as distinct from
traits, must also involve time . Failure so to do renders studies of moods and
states of little value, in his view, which is certainly logically impeccable.
This means that regular R factor analysis involving correlations between
tests, the usual approach to the field (other than by Cattell), must be treated
with considerable caution because the factors could be traits or states.
Howarth (1980) is a good example of a careful R factor analysis of
moods, which must be scrutinised. He based his mood adjective checklist
upon a study of all previous mood scales and the R factor analysis revealed
the following moods: a~'l"ession, scepticism, egotism, outgoingness, con-
trol, anxiety, cooperation, fatigue, concentration and sadness. As has been
mentioned, their validity as states is dubious because they emerged from
R analysis. Ifind it difficult to consider scepticism to be a state. Surely, iri
some individuals, for example Socrates or Bertrand Russell, scepticism was

more than this.
Of course, workers in the field of moods who use R technique try to
counter the criticism of Cattell by differentiating states and traits in terms
of items. State items stress the present. 'I feel sad now' is a state item while
'I generally feel sad' measures the trait. This gives considerable face validity
to mood items of this kind. Nevertheless it is not a satisfactory manoeuvre.
Thus a person who is generally sad will be sad when tested.
Recently, despite these difficulties with R analysis, there has been a
renewed factor analytic attack on the problem of the structure of moods
by Tellegen and his colleagues. Watson and Tellegen {1985) surveyed a
large number of studies of mood and reported some of their own studies.
They concluded that two orthogonal factors would account for much of
the variance in mood scales: positive affect and negative affect. All moods
and states which are pleasant load the first factor and all unpleasant states
the second. This is a powerful simplification of the field which essentially
states that there are only two moods, pleasant and unpleasant.
Cooper and McConnille (1989) factored the Cattell Eight-State Ques-
tionnaire (which is discussed below) and the Tellegen scales. They showed
that state anxiety was equivalent to negative affect and state exvia to positive
affect. This suggests that the Tellegen scales (despite the R analysis) are
equivalent to the second-order factors in the Cattell set (see below).
Cooper and McConnille (1990) also examined mood variability using
again the Eight-State Questionnaire (Curran and Cattell, 1974). They
found that there were considerable individual differences in mood vari-
ability. Thus subjects who varied a lot on one of the scales tended to vary
a lot on all of them and vice-versa. Such results are explicable in terms of
common factors underlying the Cattell scales and the Tellegen positive and
negative affect factors could well be the ones.
Such variability has certain clear implications. Norms are not appropri-
ate given the variability in scores, although norms for the actual variability
itself might be possible. Single measurements of variables which fluctuate
are only useful if the measurements are required at that time. Finally, if a
general level of a subject's mood were to be needed several measurements
are required.
The implications of all these findings from the admittedly somewhat
dubious R factor analysis of states and moods will be discussed after the
work of Cattell has been scrutinised.
Work of Cattell on moods and states

As has been mentioned Cattell has criticised R analyses as being unsuited
to the analysis of moods. However, as was pointed out in Chapter 3, there
are other forms of factor analysis than the usual analysis of the correlations
between tests. These will be discussed below.
P technique
In P technique the correlations between occasions on a variety of tests for
each individual are subjected to factor analysis. Thus these factors account
for variance over time within an individual and must be moods or states.
P factors could never be traits. However, P factor analysis has not been
much used on account of a number of problems.
Since each individual has to be tested many times on each variable it is
difficult to obtain subjects who are willing to do the tests. Even payment
may not be sufficient to produce the required dedication.
2 Constant retesting is highly likely to affect the validity of tests which
are not designed for such use. Subjects get bored doing the same items
or remember what they put before. Sometimes semantic satiation occurs
when the items are so familiar that they lose all meaning, a phenomenon
which can be experienced by continuously repeating a word.
3 Time intervals. Since some moods or states last only a few minutes, time
intervals to catch these would have to be impossibly short. Even testing
everyday will miss some moods.
4 Sampling problems. Not only is it difficult to obtain many subjects for
P studies, it is dear that those who are willing are unlikely to be a
representative normal sample.
5 Case studies. In P technique the factors obtained are unique to the
individual from whom they were obtained. This means that at the end
of a research a collection of case studies has to be interpreted. Given all
the other problems this has led Cattell (1973) and Cattell and Kline
(1977) to propose two other possible methods.
dR technique
This is an R factor analysis of the differences in scores on tests of subjects
on two occasions. This approach allows the use oflarge and representative
samples and the factors, this time common to all subjects, must be moods
or states because they account for variance over time. Again, as in P
technique, they cannot be traits.
However, Cronbach (1984) has criticised the use of difference scores as

a basis for statistical analysis on account of their large standard erron, a
particularly serious problem with factor analysis. One way round this is to
split the sample and to interpret only replicable facton.
Chain P technique
In this the attempt is made to combine the advantages of both methods.
Thus iftwenty subjects are tested on five occasions chain P technique treats
the data effectively as if there were 100 testings. However, as Cattell (1973)
argues this is a compromise and P technique is to be preferred.
Conclusions from the discussion of methods From this discussion

of methods in the elucidation of states a number of conclusions can be
drawn. The point noted previously concerning the relativity of stablility
and transience, as applied to states and traits, must also be borne in mind.
1 Traits can only appear in R analysis.
2 States can appear in R, dR and P analysis.
3 Change facton can appear in dR and P analysis.
Here a new term, change facton, has been used and this must be defined.
In the initial studies of moods and states reported in Cattell (1973) and
Cattell and Kline (1977) a number of facton emerged from dR analyses
which were similar to the traits and it made no sense to think of these as
states or moods. Cattell labelled them trait-change facton. These represent
the growth and decline of traits and are not motivational or moods or states.
However, in all studies of states based on dR analyses or P technique it is
always possible that trait-change factors may arise and they have to be
distinguished from genuine states.
There are two interesting examples here- anxiety and extravenion. As
has been made clear in Chapter 4, anxiety and extravenion are the two
largest personality facton. In dR studies both of these emerge. Are these
therefore trait-change or state facton? Anxiety is usually considered to be
a state in most theories of personality and in common experience. Thus
few would attempt to argue that the anxiety of dR analysis is a trait-change
factor. However, exvia (Cattell's equivalent to extravenion) also appears
in dR analyses. This is not normally conceptualised as a motivational or
state factor so that it would, at fint glance, appear to be a clear trait-change
factor. However, Cattell argues that this may be a genuine state factor since,
as we all know from experience, on occasions we feel sociable and on
others we cannot face even a brief conversation.
Conclusions
From this discussion it is clear that, to some extent, the distinction between
trait-change and state factors is subjective and that it is impossible to claim
that states discovered only through R analyses are necessarily states. Factors,
on the other hand, revealed through P and dR analyses must be states or
trait-change factors.
Twelve Cattell state factors

From this dreadful and complex morass Cattell has argued that there are
twelve state factors, although their validity and general support from other
empirical work is far less than that with the personality traits. The twelve
factors are: anxiety, exvia, cortertia (alertness), independence, depression
(general), psychoticism, stress, fatigue, arousal, regression, and two minor
depression factors.
Two tests have been developed to measure the largest of these factors,
the Eight-State Questionnaire (Curran and Cattell, 1974) and the Central
Trait-State Kit {Barton and Cattell, 1981) which measures the five largest
second-order states and traits in the Cattell system.
Final conclusions concerning the factor analysis of moods and states

The factor analysis of moods and states has not yet reached a generally
agreed structure, partly because most workers are happy with R analysis
despite its problems, while only Cattell has applied P and dR analysis.
Nevertheless as the study of the Tellegen scales (the best R scale) and the
Eight-State Questionnaire {the best dR scale) by Cooper and McConnille
(1989) showed, there is considerable overlap. State anxiety was equivalent
to the negative affect scale, state extraversion to the positive affect. These
authors also showed that individuals could be described along a dimension
of mood variability, which supports the claim of common factors under-
lying separate moods.
Certainly to conceive of only two moods, pleasant and unpleasant,
whose outward manifestations may vary according to circumstance, is an
elegant account of moods and states but further research is needed to
indicate whether this simplification is not too drastic.
It might perhaps be worthy of note, that the fact that anxiety and exvia
can be seen as states (change factors) throws some doubt on the distinction
between state and trait. It could be argued that certain traits such as anxiety
or obsessionality, as was suggested on general principles, at the beginning
of this chapter, also act as drives and this accounts for the emergence of
trait-change factors.
Factor analysis of drives: motivational structure

The factor analysis of drives has proved to be even more difficult than that
of states and moods. Since Cattell is about the only psychometrist who has
attempted to factor drives, this section of this chapter will be largely
concerned with his work, of which good accounts can be found in Cattell
(1985) and Cattell and johnson (1986).
Definition of drives
Cattell sees three aspects to drives, in the tradition of McDougall (1932).
There is a tendency to attend to certain stimuli rather than others, e.g. food
when hungry rather than flowers. Each drive has its own characteristic
emotion and there is an impulse to some particular course of action.
Clearly, therefore, to understand motivated behaviour in these terms it is
necessary to elucidate the structure of drives which underlies it.
Drives can be looked at in another way: as Cattell (1957, 1985) argues,
attitudes reflect drives because the strength of an attitude reflects the
strength of an impulse to action in response to a stimulus. It is on these
grounds that Cattell further argues that the factor analysis of attitudes will
reveal drives. This conceptualisation of drives also implicates, it should be
noted, interests. These also must be held to be accounted for by drives.
This is certainly true incidentally in "psychoanalytic theory where, for
example, interest in surgery reflects sublimated aggression and in art,
sublimated anal erotism (Fenichel, 1945). Hence the factor analysis and
measurement of interest ought to reveal motivational factors.
Measurement of interests
At this point a brief paragraph on the measurement ofinterests is required.
This is because there is a considerable number of interest tests which have
been largely developed for practical occupational psychology to aid in
selection. and career development, a use of psychometrics which is dis-
cussed in Chapter 8 of this book. However, such tests have revealed
nothing concerning the nature of human motivation because they were
constructed not by factor analysis which might have revealed underlying
determining factors, but, in the main, by criterion-keyed methods which,
as was shown in Chapter 2, yield variables of unknown psychological
meaning. I shall briefly mention these and other non-factor analytic tests
of motivation at the end of this chapter but essentially they are unable to
yield useful information about dynamic structure.
Strength of interest
There is one further aspect to interests which needs to be measured. This
is strength of interest. It is obvious from experience that not only do
interests differ but so also does strength ofinterest. Cattell and Child (1975)
list from a search of the literature relating to expressions of attitude and
interest sixty-eight indices and report the results of a number of simple
structure factor analyses. Such tests as these are, in the categories adopted
in Chapter 2, objective tests. These strength of interest factors are set out
below:
Alpha- 'conscious id' This is the component related to the satisfaction

of personal desires even when this is unwise. Politicians caught in sexual
or financial scandals exemplify the working of this component.
Beta - realised integrated interest This is the ego component of

attitudes, the rational aspect, the only part tested by standard attitude
questionnaires.
Gamma- 'superego' The moral component of interests, the aspect of

interest revealed in the pseudo-interests of the middle classes in the arts.
Delta This is a physiological component to interest reflected in the

thrilling sensations - spine tingling - to certain stimuli.
Epsilon This seems to be a conflict factor, perhaps related to depression.

There were two further unidentified primary factors.
At the second-order, three factors emerged:
Integrated component This loads on beta and gamma and reflects

reality and information-based experience - the rational aspect of attitudes
and interest.
Unintegrated component This loads on alpha, delta and epsilon and

thus reflects aspects of interests below the level of awareness.
There was a third unidentified component.
As has been argued by Kline (1981) these strength of interest factors,

especially at the second-order, fit a general psychoanalytic model of
interests in that their strength reflects conscious and unconscious compo-
nents.
These findings have important implications for the measurement and
understanding of attitudes, interests and motivation. Thus it is clear that all
attitude measurement must involve these factors, certainly the two main
second-orders. As has been mentioned, the standard attitude questionnaire
which contains simple face-valid items relevant to the interest or attitude
cannot do this. Such scales are solely concerned with the integrated,
rational component. This almost certainly accounts for the weak predictive
power ofstandard attitude scales. Cattell (Cattell, Horn and Sweney, 1970)
has developed the Motivational Analysis Test to measure the most import-
ant drives as well as interest strength and this will be discussed later in this
chapter.
It must be pointed out, however, that beyond the boundaries of Cattell
and his colleagues, this factor analytic approach to attitudes and motivation
is not only not accepted, it is virtually unknown. This is particularly
unfortunate since it means that the factors have been little explored so that
there is little external evidence to support them.
In brief these strength of interest factors still require further research
before they could be accepted as definitive accounts ofthe attitude strength.
However, they represent psychometric hypotheses which deserve proper
exploration.
The structure of drives

The same objective tests, based on indices and expressions of interest, can
also be used to tap drives and this has been done by Cattell and his
colleagues and reported in Cattell and Child (1975) and Cattell (1985).
Although the work of Cattell is empirical, there were models and
theories broadly underlying the selection of tests in the first place. A truly
atheoretical collection of data is impossible. These theories were broadly
those of McDougall (1932), Murray (1938) and Freud (1933), who have
in common the notion that drives are reflected in interests and that various
behaviours can be traced back to certain goals and that one behaviour may
relate to several goals. For example, an interest in cars may reflect various
goals: to arouse envy, to feel powerful, to attract women or men, to exhibit
wealth.
This raises yet another aspect of the structure of drives. In analysing the
goals of any behaviour, it is often found that the goals are ordered, until
ultimately a final goal is reached. Examples of these are to get food, to

obtain warmth, to enjoy sexual activity. These are considered by Cattell
and most other theorists to be basic biological drives, which can be
observed in many other organisms, especially mammals. In Cattell's
neology these basic biological drives are ergs. These are contrasted with
sentiments which are culturally moulded drives, uniquely human.
As has been mentioned, Cattell and his colleagues have subjected
objective tests which, from studies of the psychology of motivation, would
appear to reflect drives, to simple structure factor analyses and the most
recent list of ergs and sentiments is set out below. Further details of these
factors may be found in Sweney et al. (1986) and Gorsuch (1986).
A Replicated ergs:
Food-seeking; escape to security; mating; self-assertion; gregarious-
ness; narcissism; parental pity; pugnacity; exploration; acquis-
itiveness.
B Other ergs (of uncertain identification):
Appeal; constructiveness; rest-seeking; self-abasement; and (even less
clearly defined) laughter and disb'liSt.
C Well-defined sentiments:
Career; self-sentiment; home-parental; sports and games; mechan-
ical; partner; religious; superego.
I shall discuss the ergs first. As was the case with the strength of interest
factors, these drive factors have been little investigated except by Cattell
and colleagues which inevitably means that there is still insufficient
evidence to establish their validity. Far more research than is reported in
Cattell (1985) or by Sweney et al. (1986) is required before it is possible to
regard these as established motivational factors. Again as was the case with
the strength of interest factors this is particularly unfortunate, since this list
of ergs, ifshown to be correct, would constitute an important contribution
from psychometrics to personality research. This is because it demonstrates
that there are more drives than postulated by Freud and less than the vast
lists of Murray and McDougall. In other words, psychometric measure-
ment has been able to improve on the speculations of even the best
theorists. Thus it is essential that efforts are made to validate and explicate
these factors .
As regards the list ofsentiments, the position is different. Although these
are the best-defined sentiments, Gorsuch (1986) admits that this is not a
complete or definitive list. Nevertheless, as can be seen, they do represent
the most obviously important interests and drives in Western culture.
These are oflittle theoretical psychological interest but represent activities
which happen to be important in that culture. Indeed these sentiments are

not dissimilar to the variables of interest tests completed by item analyses
of face-valid items derived from common-sense knowledge of human
interests and activities, tests which will be discussed at the end of the
chapter.
As was pointed out above, Cattell has developed one objective moti-
vation test, the MAT (Cattell, Hom and Sweney, 1970), based upon his
factor analytic studies of motivation, and this will be described. I shall
further examine a factor analytic measure of interests, designed specifically
for occupational psychology, the Vocational Interest Measure, the VIM
(Sweney and Cattell, 1980).
The Motivational Analysis Test

The MAT (Cattell, Hom and Sweney, 1970) is one of the few objective
tests which has been published as a test for general use. This measures the
ten best-established motivational factors - five ergs and five sentiments.
These are: mating, assertiveness, fear, narcissism and pugnacity (the ergs);
self-sentiment, superego or conscience, career, partner and parental home
(the sentiments). In addition the integrated and unintegrated components
of interest strength are measured.
There are four kinds of objective test in the MAT:
1 Questionnaire items in which subjects have to indicate what is the better
use of a given amount of time, money or some other commodity.
2 Estimates of feelings have to be given on a four-point scale. For
example, 'What percent of adults are happy to give to charity?'
Both these tests measure the unintegrated component of drives and it
should be noted that, although they are of the questionnaire form, they are
objective because the purpose of the items is hidden from subjects including
most psychologists.
3 Paired words in which subjects have to indicate which of a pair ofwords
goes better with a key word.
4 Information. This is a knowledge test, claimed by Cattell and Child
(1975) to measure interests because people know most about what they
are interested in.
Both these tests measure the integrated, rational component ofstrength
of interest.
Reliability and validity of the MAT There is little doubt that the
Cattellian approach to the measurement of motivation is brilliant in

conception. However, its instantiation in the MAT is not satisfactory. In
the handbook to the test the scales are oflow reliability, the median alpha
being only 0.45.
Kline and Grindley (1974) showed, in a twenty-eight-day case study
where a subject completed a diary and the MAT every day, that fluctuations
in ergs and sentiments corresponded to diary events in a striking manner.
This led us to investigate the validity of the test with more psychometric
rigour.
To this end Cooper and Kline (1982) carried out an oblique simple
structure rotation of the 16PF and MAT but only eight factors emerged
and none of these was in accord with expectation. Item analyses also
showed that the scales were not homogeneous, items not fitting the scales
of which they were a part. From this study it must be concluded that the
MAT is not a valid test, at least in Great Britain, although it certainly
deserves much further research and development.
The Vocational Interest Measure, VIM (Sweney and Cattell, 1980)
This,like the MAT, is an objective, factor analytic test, but one specifically
designed to measure occupational interests and to be useful for occupational
psychology.
Variables measured Two ergs: protectiveness and rest-seeking. Eight

sentiments: career, mechanical interests, clerical work interests, scientific
work interests, aesthetic-dramatic interests, business-economic interests,
sports interests and nature-outdoor interests.
Although first developed in 1980, this was an experimental version of
the test and, as the article by Sweney et al. (1986) shows, at present it still
requires external validation against success at relevant occupations. With-
out this it cannot be used for anything but research into its own validity
and into the viability of Cattell's factor analytic approach to motivation.
The interest of this test lies in the fact that it is a part of a factor analytic
exploration of motivation, although it is likely that the variables which the
VIM attempts to measure are ofless psychological importance than those
in the MAT, because these seem to be fundamental aspects of the human
organism (ergs) and of human culture. The VIM variables, on the other
hand, seem applicable only to a particular set of jobs in Western society
and although the test may prove valuable in the practice of occupational
psychology it seems unlikely to throw much light on motivation in general.
Before completing this chapter by drawing out substantive conclusions
from the psychometric study of personality, brief mention must be made

ofmotivation and interest tests which were not derived from factor analysis.
Two of the most widely used American tests, the Strong Vocational
Interest Blank (Strong et al., 1971) and its modem form the Strong-
Campbell Interest Inventory (Strong and Campbell, 1974), and the various
forms of the Kuder Tests (Kuder, 1970a, 1970b), for a variety of reasons
can make no strong contribution to the psychometry of motivation. Thus
the Strong tests were developed by criterion-keying and items were
selected which could discriminate one occupational group from another.
Such scales, as has been argued in Chapter 2, have no psychological
meaning m that the results are only useful if good discriminations can be
made between groups. In fact the discrimination is little better than
expressed interest.
The Kuder Occupational Interest Survey (Kuder, 1970b) in which the
items were selected if endorsed by an occupational group, an alternative
form of criterion-keying, again has scales which lack psychological
meaning. Furthermore, this method ensures that many individuals seem
suited to a variety of occupations, which is useful for discussion and
counselling but of little practical value in selection. Another form of this
test, the Kuder General Interest Survey (Kuder, 1970a), is ipsatively scored
which makes comparison between the scores of individuals meaningless.
While this test may be useful for discussion in occupational counselling it
cannot make any contribution to a knowledge of motivation.
The last test which I shall discuss is the Vocational Preference Inventory,
the VPI, constructed by Holland (1985a) and the measuring instrument
developed as part ofhis theory of vocational choice (Holland, 1985b). This
test, in as much as it is part of a theory, is of potentially considerable
psychological interest if the validity of either test or theory could be
supported.
Description of the VPJ

The items ofthis test consist ofoccupational titles to which subjects indicate
like or dislike. These form nine interest scales: realistic, investigative,
artistic, social, enterprising, conventional, self-control, masculinity-femi-
ninity and status. The internal consistencies of these scales are generally
high although three are only around 0.5.
Over more than thirty years Holland has developed his theory of
occupational choice and the results of more than 400 investigations with
the VPI can be called upon as evidence. Nevertheless, despite this huge
array of results, the research can be summarised.
Holland (1985a, 1985b) has shown that there is a moderate correlation

between his personality types as defined by the scales and job choice. This
supports Holland's occupational theory which claims that personality types
seek out jobs suited to them. The predictive and concurrent validity of the
VPI scales relative to job choice is at least as good as that of the Strong and
Kuder tests.
Holland also regards these scales as measures of personality and has
correlated them with the main personality test scales discussed in Chapters
4 and 5. There are numerous moderate correlations which make reasonable
psychological sense. Thus, for example, the VPI enterprising scale corre-
lated with Cattell's A, sociability; E, dominance; F, enthusiasm; and H,
adventurousness. All this suggests that the VPI may be measuring nothing
more than the ubiquitous big five factors or even more simply the two
largest of these, extraversion and neuroticism. This means, of course, that
strictly the VPI is a temperamental rather than a dynamic scale.
To some extent this is supported by the study by Costa et al. (1984)
with the NEO inventory when one scale correlated with neuroticism, two
with extraversion and three with openness. An ongoing study in which
the 16PF and the VPI were subjected to oblique, simple structure factor
analysis strongly suggests that two factors can explain the VPI variance -
neuroticism and extraversion - and that these are more associated with
career choice among adolescents than are the simple VPI scales (Parker and
Kline, 1992).
Conclusions concerning the VPI This test which is tied to the theory
of Holland concerning occupational choice is probably no worse than
many others at predicting choice. However, this is done with simple
face-valid items and is of little psycholob>'ical interest. More interesting is
the fact that these scales probably together measure anxiety or neuroticism
and extraversion and are not independent. It is clear no powerful motiva-
tional theory could depend on the VPI.
Final conclusions
From this chapter it is quite clear that factor analysis has not so far led to
considerable substantive conclusions. This is because the majority of users
of motivational and interest tests are applied psychologists who are not
concerned with theoretical issues such as the number of drives or the nature
of drive strength. All they require is an effective test for selection and
counselling. Such tests can be easily made by techniques which do not lead
on to theoretical insights.
The only factor analytic attempt of any substance in the motivational

field is that of Cattell but, as has been seen, there are profound difficulties
in establishing mood or state factors (where R analysis is insufficient) as
well as other dynamic factors.
However, as a result of the reluctance of practical occupational psycho-
logists to use these factored tests, the factors emerging from the Cattell
work have not been validated in the field so that there is insufficient
external evidence relevant to their psychological meaning. Furthermore,
factored studies of the tests do not suggest that they are measuring the
variables claimed for them.
In brief, this is a field in which psychometrics has not yet made a
substantive contribution, except perhaps in the notion of only two state
factors - positive and negative affect, although even here these factors
require further validation. However, the methods and approaches, P and
dR analyses and the validation of emerging factors, are all ready for use
and all that is needed is technically adequate research.
Chapter 7
Heritability of personality
The question of the heritability of personality is one of the most fundamen-

tal in the psychology of personality. This is because any psychological
theory must take into account the heritability of the variables with which
it is concerned. Thus if, for example, the heritability of anxiety were zero
it is clear that its status would be entirely dependent on environmental
factors and any theory must be able to state what these are and how they
affect anxiety. Similarly if environmental effects are zero, as in the case of
eye colour, to postulate environmental influences must be wrong. Clearly
then, to determine the heritability of personality variables is essential for
any adequate theory of personality and for a good understanding of the
field.
In addition, there are further powerful arguments from the study of the
heritability of personality which can be brought to bear on the problems
of understanding personality. The first relates to already extant theories of
personality which are either refuted or supported by the findings. Thus if
it turns out that much of the variance of personality is determined by
genetic factors, theories in the psychoanalytic tradition, in which great
emphasis is placed on the importance of child-rearing, are in some
difficulty.
The second argument is only relevant to the psychometric approach to
personality but is of critical importance to the interpretation of the
personality factors which have been discussed in the last three chapters. As
was made clear in our discussions in earlier chapters of test validity and of
factor analysis, a major problem in any psychometric factor analytic study
is the identification of the factors or their validity. Normally they are
interpreted from their factor loadings and by external validation, although
this latter is often difficult to establish in the case of personality factors.
Establishing the heritability of personality factors is a particularly useful
98 Heritability of personality
form of external validation because if it turns out that a factor h2s

considerable genetic determination, a major criticism of factor analytic
personality variables is, at a stroke, removed. This criticism is the always
logically possible argument (in the case of personality questionnaires) that
the factors are no more than bloated specifics, collections of items that are
essentially paraphrases of each other. It makes no sense that a putative
bloated specific has a considerable portion of its variance determined by
hereditary factors, so that, if it thus turns out, the factor cannot be specific.
Factors whose variance is determined to any considerable extent (and this
will be precisely defined later in this chapter) by hereditary factors cannot
be any kind of statistical artifact and must be of some psychological
significance.
In brief, determining the heritability of personality variables is of
considerable theoretical interest in the study of personality, and is perhaps
especially important to the psychometric factor analytic view of person-
ality.
Biometric analysis
In this chapter I shall examine the findings from the biometric analysis of
personality. Biometric analysis involves the study of the sources of variance
within populations with reference to genetic and environmental compon-
ents. Several important points about biometric analyses should be noted at
the outset before a more detailed description and rationale of the proce-
dures is given.
Within populations This is a highly significant phrase. If it is found, for

example, that 40 per cent of the variance of a particular personality trait is
determined by genetic factors it does not mean that 40 per cent of that trait
in any individual is determined genetically and the rest by her environment.
These figures refer to sources of variation within a population.
It is also important to note that the heritability ratios computed for one
population are not necessarily the same for another. Thus in a population,
for example, where environmental factors were oflow variance (perhaps
where there were strict rules of child rearing), hereditary factors would be
relatively of greater significance in determining individual differences.
Finally at this juncture, it is important to remember that if within a
population a trait has a considerable genetic determination, it does not
follow that differences in that trait between two populations must be so
determined, as Plomin (1986) has demonstrated.
Heritability of personality 99
Biometric methods
Although these methods are algebraically complex, Fulker (1979) has
provided a brilliant, simplified version which I set out below.
P = G + E where P is the phenotypic variance (the observed variance,
e.g. the scores on a test}, G is the variance determined genetically and E is
the variance environmentally determined. By the use of variances in this
model it is possible to separate out G and E from the variances and
covariances of groups of individuals such as twins.
The analysis of variance of twin pairs partitions the variance into two
sources: between and within pairs. The more pairs resemble each other,
the greater the between pairs variance will be compared with the within
pairs variance. Indeed, the ratio of (B-W)/(B+W) yields the intra-class
correlation showing how similar pairs of twins are. From these variances
and correlations, the genetic and environmental components can be
derived.
E, the environmental variance, can be broken down into two parts: the
common or shared environment (CE) reflecting the experiences of home
life which are common to members of a family and the specific environ-
ment (SE) or unshared aspects of experience. The biometric equation can
then be written:
P= G + CE + SE.
With this model a number of deductions can be made.
The correlation (r) reflects the variance of all shared influences.
a Identical (MZ) twins. r = G + CE;
b Non-identical twins. r + Y2G + CE (these having half their genes in
common).
From these assumptions the following estimates can be made:
G = twice the difference between the two correlations;
CE = the difference between the MZ correlation and the estimate of
G;
SE = 100-G + CE.
This is the basic reasoning behind the biometric approach to the analysis
of the genetic and environmental determinants of the variance of any trait
within a population. It should be noted that this is the most simple additive
model. However, it is possible to increase its complexity and its accuracy
by using the intra-class correlations derived from relatives other than twins,
and allowing for dominance and assortative mating for example, where this
occun, as it does in the case of intelligence where more intelligent

individuals tend to choose more intelligent partnen (and have more
intelligent children).
From this discussion it can be seen that biometric analyses determine
the sources of variation of traits within populations by examining the
variances and covariances among individuals of all degrees of relatedness
reared apart and together. Relatives reared apart are a particularly valuable
group because this sets CE in the equations at zero.
AsJinks and Fulker (1970) argue, the fact that these biometrical methods
were developed for agricultural and biological use is no reason for suppos-
ing that they would not work in the human case. All that is required for
these methods are trait measures and the requisite samples of relatives. This
is what makes biometric methods well suited to the psychometric approach
to penonality since this involves penonality tests yielding numerical data.
From our description of biometric methods it is dear that twin studies
are important to them. However, on their own, although they have been
traditionally used in investigations of the influence of heredity, especially
of intelligence, they are inadequate for a number of reasons. Fint, studies
of identical (monozygotic) twins reared apart are likely to underestimate
the contribution of genetic factors. This is because there are sources of
discordance among identical twins which, although neither environmental
nor genetic in the ordinary sense of the words, would be classified as
environmental in twin studies where all differences must be so attributed.
These special facton which have been described by Darlington (1970)
include nuclear differences which arise by gene mutation or by chromo-
somal loss or gain when the zygote splits; cytoplasmic differences which
are brought about by deleterious genes acting differentially on the two
organisms; embryological differences created by erron arising in a late
splitting and what is perhaps the most common: nutritional differences due
to unequal placentation. This last is probably highly important in the light
of recent research which shows that early nutrition (profoundly affected
by placental size) has considerable physical effects in twins {Bryan, 1992).
There are further problems with twin studies. For example, the number
of twins reared apart is obviously small and they may not be a representative
sample of identical twins, thus making generalisations to normally reared
twins difficult. This is an extension of the problem of whether it is safe to
extrapolate from twins to singletons. The twin method assumes that genetic
and environmental influences are uncorrelated and that they combine
additively without interaction. This assumption may be difficult since it is
likely that the environ_ment of intelligent individuals, for example, is
somewhat different from that of the less intelligent. Intelligent children
respond differently than the less intelligent and provoke different responses.
Finally, the twin method can be criticised on the grounds that twins are
treated more alike than are singletons, especially MZ twins, thus apparently
increasing genetic determinants of similarity. However, Loehlin and Ni-
chols (1976) have shown that the personality scores of pairs of twins treated
similarly were no more alike than the scores of twins not so treated,
although the work ofRose et al. (1988), which is discussed below, indicates
that increased social contact may enhance similarity.
However, in contrast to working with simple differences between
twins, in biometric methods, as has been shown, the total variance is broken
down into the between families and within families variance. In addition,
the contribution of the interaction of the genetic and environmental
variance, as well as the correlation between them, can be taken into
account. Indeed, as Jinks and Fulker (1970) argue, one of the great
advantages ofbiometric analyses is that it is possible to test different genetic
models. There is no need to postulate a simple linear model but the effects
of correlated genetic and environmental factors and of interactions can be
investigated as well as the influence of gene dominance and assortative
mating.
It seems difficult to attack this biometric approach to the investigation
of genetic factors in personality particularly where tllesurdies use the more
complex models discussed above, since in the fields of biology and
agriculture it has proved highly valuable. However, Feldman and Lewontin
(1975) have argued that analysis ofvariance cannot separate variation which
results from environmental fluctuation from that due to genetic segrega-
tion. However, it appears that these arguments do not apply to the
biometric analyses discussed by Jinks and Fulker (1970) and the subject of
this chapter. Thus Feldman and Lewontin (1975) claim, and this is one of
their strongest objections, that broad heritability, total genetic variance, is
not a useful statistic in human population genetics. What is important, they
argue, is the narrow heritability, the proportion of variance due to additive
genetic variance. However, in the biometric analyses in this chapter, both
broad and narrow heritability can be computed, and it is clear that these
criticisms are not pertinent to these procedures.
One further possible objection remains to be discussed. Uninformed
criticism of the findings ofbiometric analyses in terms of variance attributed
to genetic or environmental factors has often been concerned with the
environmental findings . Such critics argue that since measuring the envir-
onment is so difficult, partly due to the fact that it is not an objective but
a subjective phenomenon, and indeed there are no well-accepted measures
of environmental variables, then the results must be flawed; poor measure-
ment leading to inevitable error. This however is mistaken, since estimates

of the environmental component of population variance are derived by
including in them what cannot be attributed to genetic factors. This is clear
from the equations at the beginning of this chapter.
The MA VA method
This is the method used by Cattell to compute the heritabilities of the
factors in his system, factors which have been described in earlier chapters
of this book. I shall describe it briefly because some results derived from it
will be discussed later in this chapter. Cattell (1982) in a detailed account
of his work on the inheritance of personality and ability sets out the
complex algebra of the MAVA (multiple analysis of variance) method for
computing the heritability of traits. This MAVA method deals with four
sources of variance: variability among siblings due to within family genetic
variance; variance within the family due to environmental influences;
variance between families due to genetic influences and variance between
families due to environmental differences. In addition, covariances and
interaction terms can be added. To solve these equations data are needed
from twins, identical and non-identical, siblings and individuals with other
degrees of relationship reared up together and apart.
I shall not describe the MAVA method in more detail because, although
it is undoubtedly superlative, Jinks and Fulker (1970) indeed describe it as
a brilliant one-man attempt to develop a statistics of genetic biometrics, it
has certain difficulties which the more usual biometric analyses described
earlier have overcome. One difficulty with MAVA is that a subjective
decision has to be made as to whether interaction terms are included in
the equations or not. This is because to include all possibilities would
involve impossibly large and usually unobtainable samples. Furthermore
there are some doubts about the logic of the algebra of MAVA, although
this is a topic which cannot be dealt with here. Given these problems, it
seems more sensible to deal with the standard biometrical analyses rather
than the more idiosyncratic methods of Cattell. However, in most cases,
MAVA and the standard procedures are in good agreement where both
have been used as in the case ofintelligence (see Cattell, 1982). Thus where
only MAVA has been used I shall treat the results as worthy of brief
discussion although some caution needs to be shown on account of the
problems which have been raised above.
Findings from the biometrical analyses ofpersonality

In the remainder of the chapter I shall discuss the heritability ratios and the
models of genetic and environmental variance which have been deter-
mined for the personality factors shown to be important in the last three
chapters of this book. As has been indicated, the results to be discussed
were obtained either from orthodox biometric analysis of personality test
scores or from the MAVA method in the case of Cattell's factors.
Much of this discussion will turn on the heritability of the Eysenck
factors because, as has been seen, these are three factors which account for
much of the variance within personality questionnaires and whose psycho-
logical meaning is known. For this purpose there are two excellent sources
ofinformation- Eaves et al. (1989) and Eysenck (1990)- and readers should
consult these for further details. As this is written Oune, 1992) an excellent
new and clearly written book is about to appear (Loehlin, 1992) which
gives a fine exposition of biometric approaches and summarises much of
the evidence for personality tests.
As Eysenck (1990) points out, there are six modern investigations of the
genetics of personality, which employ biometric analyses and these form
the basis of the summary reported below. These are:
1 Work on the EPQ with 500 pairs ofMZ and DZ twins, fully reported
in Eaves et al. (1989).
2 Research by Loehlin and Nichols (1976) on 8,660 sets of twins in the
USA. Unfortunately they used the CPI which is a criterion-keyed test
ofdubious validity (see Kline, 1992a) although they attempted to extract
face-valid measures ofE and N from it.
3 An Australian study by Martin and Jardine (1986) using 4,000 pairs of
twins and the EPQ plus other attitude and personality measures.
4 A Swedish study with nearly 13,000 pairs of twins and a short form of
the EPI, a precursor of the EPQ, measuring E and N (Floderus-Myrhed
et al., 1980).
5 American studies, e.g. Tellegen et al. (1988) on twins reared apart and
together. However, these authors used a test of personality of unknown
validity, thus making the results difficult to interpret.
6 More than 7,000 pairs of adult twins studied by Rose et al. (1988) in
Finland.
The main results are as follows, although it should be noted that these
biometric analyses are based upon twins.
A The between family environmental variance, the variance from the
shared environment, is a trivial component of penonality variance. This is

a finding which has been replicated in almost all studies (other than Rose
et al., 1988) and is based on data from 106 pain of identical twins reared
apart and more than 5,000 pain of identical twins reared together.
This is an astonishing finding which is contrary to almost all theories of
penonality which stress the importance of the family environment. Facton
such as the education and social class ofthe family, their attitudes and values,
even to child rearing, have no effect on penonality. What is important in
the environmental determinants of penonality are the unique experiences
ofeach individual. Note, however, that this may include the particular way
in which a mother responds to each of her children, thus not ruling out
psychoanalytic theories, but casting considerable doubt on more socially
oriented approaches. As Eysenck (1990) argues, earlier work (e.g. Shields,
1962) which had shown that twins separated earlier were more alike than
those separated later, is consonant with this position although it is a curious
finding which requires further explanation.
However, the one contradictory investigation concerning the negligible
impact of the shared environment (Rose et al., 1988) requires some further
comment. These authon found that the longer twins were together the
more alike they were, although it could be the case that similarity caused
twins to increase their social contact, rather than the other way round.
These authon attribute the failure of most statistical studies to find
differences arising from the shared environment to erron: the limited
statistical power with which twin samples test environmental sources of
variance; the unreliability of penonality measures and the allocation of
error variance to the unshared environment; the tendency for inferences
about sources of environmental variation to be based on global estimates
rather than direct measurement and the misleading estimates of shared
environment arising from the relation of MZ and DZ correlations, mis-
leading on account of differences in social contact between the two types
of twins.
This work of Rose et al. (1988) cannot be held to refute all the other
findings, but dearly some caution should be shown in making extravagant
claims that all previous penonality theories are absolutely wrong. In this
study, increased social contact was related to greater similarity within pain
of twins.
B Penonality variance appean to have a considerable genetic component

on all the traits which have been analysed. At present in all the populations
tested this is around 50 per cent. However, if the unreliability of the
measures were taken into account, this could rise to 60 per cent. What is
particularly interesting here is not only that fundamental personality

dimensions, such as E, N and P, have these large genetic components but
so too do attitude scales of the kind which most sociologists would regard
as simply the results of social factors. Authoritarian and co~ervative
attitudes are examples of these. The work of Eaves et al. (1989) is
particularly interesting here because even at the item level strong genetic
determination can be shown, as for example the following items:
Sex crimes, such as rape and attacks on children, deserve more than mere
imprisonment: such criminals ought to be flogged or worse.
Men and women have the right to find out whether they are suited
sexually before marriage.
The average man 'can lead a good enough life without religion.
Once again, as was the case with the finding that the shared environment
was of trivial importance, the fact that such basic social attitudes are
considerably genetically determined is surprising, at the very least.
C For N the data support an additive genetic model whereas for E the
pattern is different. It follows a competition model in which a sociable
child gets the friends and leaves the books for her introverted sibling (Eaves
et al., 1989).
D In the large sample studies (Sweden and Australia) there appear to be

differences between the sexes. Thus for N there are genetic effects that are
specific to males and females. However, it must be remembered that with
large samples psychologically unimportant differences may be statistically
significant.
E Eysenck (1990) argues that there is evidence that different genes

operate at different stages of development (young and old). However,
during adult life the same genes seem to be operating (Eaves et al., 1989).
F As regards assortative mating, it does not occur with respect to E and

N and only to a slight extent in the case of P. In this, these personality
variables are completely different from intelligence where assortative
mating is important in explaining phenotypic variance.
These are the main findings from the large-scale studies of twins which
were cited above. Of these the two most striking are the importance of
genetic detenninants in the variance of the main personality factors E, N

and P and the lack of impact of the shared environment.
It should be pointed out that in a recent publication Loehlin (1992) has
summarised the biometric studies carried out with the big five factors.
These all show considerable genetic detennination, similar to that discussed
for the EPQ and again the shared environment has litde effect on the
vanance.
There are other important issues in the biometric analysis of personality
which must now be scrutinised. First, as was made clear at the beginning
of this section, these results were concerned with P, E and N although
brief mention was made of the social attitude findings of Eaves tt al. (1989)
and the fact that the big five factors appeared to be similar, in respect of
genetic and environmental determination, toE, P and N. This introduces
the important question as to what extent other personality variables are
genetically determined.
In general, studies of other personality factors have yielded similar
results, although there are differences between factors. One of the best
sources for this is Cattell (1982) in which the Cattell personality factors
were subjected to biometric analyses using his own MAVA method. This
shows an overall heritability for all traits of 0.38. However, it must be
realised that the Cattell factors are not highly reliable and this error boosts
the environmental component. Furthermore methods, such as MAVA,
which use other relatives (not only twins) tend to yield lower heritability
ratios.
Conclusions
From these studies it must be concluded that there is a considerable genetic
detennination of the population variance in personality and even social
attitudes and that the shared environment has litde effect. What matters in
personality in respect of environmental detenninants appears to be the
unique experiences of each person. Certainly theories which posit as
important in personality development variables which discriminate families
such as attitudes, social class and education, are unlikely to be correct.
Chapter 8
Personality testing in applied

psychology
In this chapter I shall discuss the use of personality tests in applied

psychology. I have chosen to examine how personality tests are applied
because these applications are the most obvious fruits of the psychometric
approach to personality, although, as will become clear, there is more to
the contribution to applied psychology from psychometric studies of
personality than good personality tests. The term 'applied psychology'
refers here to three areas in which the results of psychometric studies of
personality and personality tests are used: educational, clinical and occu-
pational psychology, each of which will be scrutinised separately although,
in fact, as shall be seen, there is some degree of overlap.
Occupational psychology
Occupational psychology is concerned with the psychological problems
involved in work. These include selection and appraisal of workers,
ensuring that workers are as efficient as possible, often done by designing
machines and organising work to be in accord with human psychology,
dealing with conflicts and disputes at work and attempting to ensure that
personnel are not discontented: stress reduction as it is sometimes called.
As might be expected from this description of occupational psychology,
the major contribution from personality testing has been to personnel
selection especially and to appraisal. It is, therefore, with these two aspects
of occupational psychology that this section of the chapter will be con-
cerned.
Personality tests in personnel selection

Psychometric testing (including also tests of ability) in personnel selection
has become big business, both in Great Britain and America, where
108 Personality testing in applied psychology
enormous numbers of psychological tests are administered for all kinds of

occupations. In this section I shall deal with the rationale and methods
rather than the detailed results which must be sought in more specialised
books (see Herriot, 1989; Kline, 1992b) andjoumals, although the main
findings, which are relatively dear, will be scrutinised.
The rationale of personality testing in personnel work

There is an assumption underlying the use of personality tests in job
selection that for any job there is an ideal specification of psychological
characteristics for it. Indeed this is an assumption of what might be called
the psychometric model of human behaviour, namely that a person's
behaviour can be predicted ifwe have a full description ofher psychological
traits. This underlies the work of Cattell (1957) who has produced a set of
specification equations for jobs (Cattell, Eber and Tatsuoko, 1970), as will
be discussed below.
It will be obvious that, with such an assumption, the task of personnel
selection becomes one of fitting people to the correct job. To such a task
there are two aspects: measuring people and measuring jobs. Obviously
personality testing is an integral part of the first of these aspects - measuring
people.
Evidence of this assumption that personality characteristics play an
important part in determining behaviour is well founded, particularly as
regards achievement and enjoyment at work. It arises from two sources.
First everyday observation strongly suggests that personality traits are
important in job success and satisfaction. Librarians, especially those
concerned with cataloguing, must be orderly and like a quiet life, quite
different it would appear from the work ofa publican or circus ring-master.
Similarly SAS officers would need somewhat different characteristics, one
would have thought, from child-care workers. These observations are
supported by the mass of empirical research which has been carried out
over the years by occupational psychologists. Huge bodies of data exist
which demonstrate that such intuitive observations are correct. There are
considerable differences in personality traits between the holders of differ-
ent jobs and personality characteristics do correlate with job success, as is
documented in Herriot (1989), Kline (1992b) and Cook (1988).
A few examples from the 16PF test (Cattell, Eber and Tatsuoko, 1970)
will illustrate this point. Artists are low on G (conscientious) and high on
M (unconventional); technical personnel are high on B (intelligence) and
low on I (tough-minded); social workers are high on A (warm) and low
on L (trusting); scientists are low on A (cold), high on B (intelligence), and
low on G and 0 (conscientiousness and guilt). These are sufficient to
indicate that personality tests are useful in personnel work. Incidentally,
Personality testing in applied psychology 109
the low conscientiousness and guilt scores of scientists are noteworthy in

the light of the need for scientists of scrupulous honesty of reporting.
Personality tests can be used in personnel selection in three ways.
Matching In this method the scores of an applicant on a personality test

are matched with the scores of the occupants of the relevant position. The
subject whose scores match best is the one selected. In a good selection
procedure, scores on ability tests would also be included in the profile of
scores to be matched but personality tests are highly useful because, as has
been seen, they usually provide several scales. Cattell, Eber and Tatsuoko
(1970) have a set of proflles on the 16PF for various occupations to which
each applicant can be statistically matched.
There are several problems with the matching method which need to
be noted.
Sampling It is essential that the sample groups to which the applicants are
matched are representative of the occupations. If they are not error is
inevitable. This means the normative groups should be large and well
sampled.
Difficulty of matching groups Since there are so many different jobs in many
cases there will be no precisely similar group to which the subjects can be
matched. In this case it is possible to match with the nearest group, but
caution must be shown in interpreting the results, for obvious reasons. A
similar point arises in the case of tests with American occupational group
norms. Even where the name of the occupation is the same it is doubtful
if the jobs require the same psychological characteristics, so that any
matching might be misleading. Indeed, in many cases, tests have so few
norms that the matching method cannot be used
In-house norms For the reasons above, personnel selectors in large organ-
isations prefer to match applicants to in-house norms. This is efficient
providing that such norms are derived from substantial samples. This, of
course, is not possible in small firms.
Changing job demands The demands of jobs, hence also the requisite
psychological characteristics, change over time, as for example when
computers begin to be used. Thus profiles for matching need to be updated
regularly.
The logic of matching A more fundamental difficulty arises when the logic
11 0 Personality testing in applied psychology
of matching profiles is examined. Clearly it assumes an ideal world where

everyone is in the job which is best suited to them. Since this is far from
the case given the pheno~enon of nepotism and the inefficiency of much
job selection, how can the matching method be justified?
Despite these problems the vast majority of job holders are reasonably
efficient and can tolerate their jobs. If this were not the case firms would
fail. Thus it is still true that a particular occupational group is far more likely
to possess the psychological characteristics necessary for that occupation
than is any other group. This is sufficient justification.
Conclusions
From this it is clear that where good norms exist the matching method is
a useful procedure. However, in practice norms may be weak or non-exist-
ent. Obviously in the latter instance the method cannot be used and in the
former modifications have to be made. Usually, in this instance in practice,
it makes sense to note the high and low scores of the relevant group and
look for similar peaks and troughs in the subjects. This makes for less
discriminating selection, for there may be many such candidates who must
then be further discriminated on other criteria. However, statistically
precise proftle matching is nonsensical unless the proftles are derived from
representative and large groups of relevant workers.
The regression method In this method regression equations, multiple

correlations, are computed between the personality test scores and job
success. These multiple correlations yield the overall correlation of the
scores with success at the job and beta weights which indicate the
importance of each score in the correlation. For example, if we were to
select lawyers by the regression method, the scores on each scale are
multiplied by the beta weights and the total ofthe weighted scores indicates
suitability for the post. The applicant with the highest weighted total is the
best suited. The higher the multiple correlation the greater reliance one
can place on the weighted total. There are several important points to be
noted about this regression method.
Sampling Clearly the samples on which the multiple correlations were

derived must be large and representative. Furthermore since beta weights
have large standard errors it is important that these weights have been
replicated in more than one study. However, correcting formulae can be
used if samples are small or if there is no replication.
Matching groups As with the first method it is essential that the groups from
whom the regression weights were obtained are relevant.
In-house norms Again, in-house norms are good provided that the num-
bers are satisfactory.
Problems in the measurement ofjob success Although it is beyond the scope

of a chapter on personality tests, the measurement ofjob success is in most
cases extremely difficult, as Ghiselli (1966) found in a study summarising
results from more than 10,000 researches. Thus it has proved extremely
difficult to produce a clear criterion ofjob success for jobs such as school
teacher or personnel worker, and this is true of any job where there is no
obvious output. However, even where there is, as say with sales persons
where volume of sales can be measured, there are difficulties. Thus, in
comparing sales of cars through a national network of dealers regional
differences, based upon affluence, the location of a dealer relative to other
rival dealers and the quality of these rival dealers, should all be taken into
account when measuring sales efficiency.
This difficulty of establishing a good criterion of success, especially where
success in a job is multidimensional, is a problem in the use of the regression
method.
Conclusions
From this it is clear that the regression method can only be used where
there are well-sampled occupational groups and where the multiple
correlations are high, say 0.7 or greater. However, provided that these
conditions are met and further that there is an adequate measure of
occupational success, the method is good and has the advantage over the
matching method that we know that success rather than simple member-
ship of a group is implicated.
In fact there are few valid personality tests with the requisite regression
equations although the 16PF test (Cattell, Eber and Tatsuoko, 1970),
despite the problems over the number of factors and despite the fact that
some of the equations use samples smaller than desirable, does provide a
set of results.
Nevertheless, it must be said about both these methods that as yet there
are insufficient test data to make them usable. Ideally as I have argued
(Kline, 1992b), an encyclopedia ofjob specifications would be established
where under each job could be found the ideal profile of psychological
traits, for the matching method, and a regression equation on to an
adequate measure of success, for the regression method. If these data were
published, occupational selection could be put onto a statistical and
rigorous basis. Since few data of this kind exist, a third method has to be
used based upon selecting the right tests.
Selecting the right tests in personnel work From the description of

the factor analysis of personality in previous chapters it is clear that four or
five variables embrace much of the personality variance - the big five -
and in any selection or appraisal procedure it would make good sense to
measure these personality factors. This is a sound fall-back procedure.
However, this is crude and even the simplest intuitive analysis of jobs
suggests that some variables would be more important than others and that
the salient variables should be measured. However, this still leaves the
question of how we know what the salient variables are. '
Task and job analysis In task and job analysis, occupations and jobs are
studied in detail to determine as precisely as possible what psychological
characteristics are required to carry them out. There are various methods
of analysis which I shall describe briefly here and for more details readers
should consult Kline (1992b) or Jewell and Siegall (1990). There are three
methods.
a Intenliews Jewell and Siegall (1990) argue that the best method of
finding out what is involved in any job is to ask those who do it. However,
the elucidation of reliable information from task analysis interviews is a
highly skilled procedure and inexperienced interviewers may emerge with
misleading information.
b Obse"'ation: 14sk analysis and description In this method a minutely

detailed description ofexactly what is entailed in carrying out a job is made.
When this has been achieved the relevant psychological characteristics are
usually obvious. Much of the difficulty in deciding intuitively what is
important in a job stems from ignorance of what it entails. This lack of
knowledge the task analysis corrects. However, adequate job descriptions
require considerable skill and expertise. It is necessary to observe the job
J>eing done and to describe it in tenns of the purpose of the whole
organisation.
Indeed, such detailed task descriptions, based upon observations, should
be contrasted with the general job descriptions based on common sense
and intuition which are often misleading. For example it is often claimed
that accountants need mathematical skills, but job analyses have revealed
that the mathematics actually used is not at all difficult. Similarly general
job descriptions may fail to reveal crisis points in procedures which detailed
analyses ought to identify. One clear example can be seen in the M1 plane
crash when an engine caught fire. The cabin crew had no emergency
procedures to report this to the flight deck and it was falsely assumed that
the pilots could see their engines.
Thus detailed observations of jobs can reveal what psychological traits
are required, although it must be remembered that such requirements are
only hypotheses which still need confirmatory evidence. Task analyses are
still relatively rare, at least in Great Britain, because they are costly and time
consuming and relatively few people have been trained to carry them out.
c Questionnaires Because of the problems involved with task analyses and

interviews, questionnaires are often used to assess the demands of different
jobs. One of the best of these is the Position Analysis Questionnaire, the
PAQ (McCormick eta/., 1972), which I shall briefly describe.
The PAQ has 194 items relating to five aspects of a job: work output,
mental processes, information input, relationships with other people, and
a general category of other characteristics. Factor analytic studies suggest
that twelve dimensions are measured by the PAQ, including variables such
as decision making, operating a machine, performing routine activities -
all dimensions, as Jewell and Siegall (1990) point out, which are somewhat
common-sense. It is a useful measure although some jobs seem not to fit
its categories and a further version has been produced for professional
occupations.
Conclusions Task analysis and description is the best method of ascer-

taining the demands of jobs. However, this is costly and requires skilled
personnel. In their absence the PAQ is better than nothing as a basis to
select tests. Ideally, tests should be selected on empirical grounds, that is
that they have been shown to correlate with success at the job or reliably
discriminate members of an occupational group. However, there is one
further point that should not be forgotten.
Use of criterion-keyed tests

As has been stressed in the discussion of personality tests throughout this
book, factored variables have psychological meaning once the factors have
been identified, whereas criterion-keyed tests, which may discriminate
occupational groups, have no necessary psychological meaning and are
usually mixtures of disparate variables. This means that if factored tests are
used in the applied field, all results have psychological meaning and
psychological knowledge can gradually be increased as more and more

results are collected. However, if criterion-keyed tests are used, there is no
accretion ofknowledge and as jobs change and new jobs come into being
the tests become useless. Thus where possible, except for mass screening,
the use of criterion-keyed tests should be avoided.
Other issues in the use of personality tests in personnel wor~

Kline (1992a, 1992b) has a detailed examination of a variety of issues
involved in the use of tests in personnel work of which the most important
will be discussed here.
Faking and deliberate distortion This is obviously a problem in all

work situations, even in appraisal, but especially selection, when subjects
know that their careers are at stake. Personality questionnaires are not
difficult to fake. Thus few serious applicants for a sales position would admit
to being shy or nervous with people. This is why Cattell and Warburton
(1967) advocate the use of objective tests which are resistant to faking,
although, as has been seen, they are not well enough validated to be used
in applied work.
Most tests have lie scales for detecting those who are distorting or faking
scores. However, all that can be done if such individuals are detected in a
selection process is to ignore their scores on the grounds that they are likely
to be inaccurate. It is not possible to correct scores for distortions.
Ipsative scores Ipsative scores, based upon forced choices, should not
be used in selection procedures. Scores of different subjects are not
comparable and thus comparisons are senseless. For this reason norms are
not meaningful even if they are supplied with ipsative tests. lpsative scores
are suited only to discussion with individual subjects and this will be dealt
with in a later section on the use of personality tests in appraisal.
Ethics and confidentiality of scores Under the Data Protection Act

if the scores of subjects are stored on computer they have the right to see
them. If they are not the question arises as to whether subjects should have
access to their scores. This is not an objective matter but one of personal
judgement.
In my view there is no reason to withhold scores from a subject. It
infringes no personal freedom of a subject to know her scores and it is
difficult to produce any argument to sustain withholding scores. A counter-
argument which is sometimes offered by practitionen is that subjects might

misundentand the results, and thus feedback might produce problems.
Computer reports from tests This argument will not do. If psycho-
logists believe that scores may be misundentood then they must explicate
them. This is a lengthy process but given the easy availability of computer
reports where large numben of candidates are involved, these should be
used. It is possible to administer questionnaires on microcomputer. The
items are presented on the computer screen and the responses are made on
the keyboard. This has the advantage that scoring can be automatic and
immediate and a print-out of the results with an interpretation can be also
immediately available. These results are based on an expert system in which
the data in the test manual are stored in the computer thus enabling
comparison with norms to be made. There are difficulties in ensuring that
the computer reports are satisfactory, especially where scores have bad
connotations, but these can be overcome with care.
It should be noted that computer-administered tests should be shown
to produce similar results to the original pencil and paper form. It should
also be realised that the results of tests administered in the standard way can
be fed into a computer and similar print-outs based on the test manuals
can be provided.
In brief, feedback should always be given to candidates after the
selection process and the use of computer reports makes this possible even
with large numben of candidates.
As a final point, it is important, also, that candidates feel confident that
the results of testing will remain confidential to those engaged in the
selection process.
Use ofpersonality tests in appraisal

I shall deal briefly with this topic because almost everything which has been
written so far about the use of penonality tests in penonnel work applies
equally to selection and appraisal. Thus in appraisal (where test results form
the basis of a discussion about individuals' career prospects) it is essential
that the tests are as valid and as relevant to the posts and as reliable as they
are in selection. The better the test the better the appraisal. However, there
are some differences, mainly of emphasis, in using tests in appraisal rather
than selection and these will be discussed below.
Validity of tests Although, as has been argued, tests should be valid, in

appraisal less valid tests can be used, provided that lack of validity is borne
in mind. Thus tests can be used as useful bases for discussion rather than
for the precision of their scores. This is certainly true of interest and
motivational tests.
lpsative scores Although ipsative scores are not useful for selection they
are valuable in appraisal. Thus the fact that a subject has ranked one score
higher than another is a legitimate basis of discussion. Appraisal is the main
arena for ipsatively scored tests.
Feedback As is obvious, appraisal is all about feedback. Essentially the

scores, where they are trustworthy, and the test items where the tests are
less efficient, are the bases of discussion rather than firm decision making.
Thus for example, if a subject is highly extraverted this can be discussed
with her and the possibility of working in a position where extravert
qualities are important can be examined.
Conclusions concerning personality tests in personnel work

From this discussion it is clear that personality tests are valuable in personnel
selection and appraisal. Ideally, factored tests should be given to good
samples of different occupational groups so that the regression and match-
ing methods can be used. In addition, task analyses and descriptions ofjobs
should be undertaken so that the requisite personality tests can be selected.
Finally the importance of feedback, especially in appraisal, was stressed.
The use ofpersonality tests in clinical and educational psychology
Clinical psychology
I shall deal first with clinical psychology, although, as shall become clear,
there is a considerable overlap with educational psychology, which is why
they appear in the same section of this chapter.
Definition of clinical psychology Clinical psychology refers to that

branch of psychology which is concerned with the treatment and under-
standing ofmental disorders. In fact this can be broken down into diagnosis,
assessment of treatment and clinical theory and personality tests have a part
to play in all of these aspects of the field.
Use of different types of personality tests

Personality inventories As was noted above with reference to occu-
pational psychology, the distinction between factored tests and criterion-

keyed tests is particularly important in clinical psychology and for much
the same reasons as were discussed previously. The great advantage of factor
analytic personality inventories is that the variables are psychologically
meaningful with the result that all findings can be used in the development
of clinical theory. In contrast, the results from criterion-keyed scales are
difficult if not impossible to interpret and psychological generalisation is
correspondingly problematic. This is particularly ironic since the most
commonly used personality test in clinical psychology is the MMPI
(Hathaway and McKinley, 1951) and its updated version MMPl-2
(Graham, 1990), both criterion-keyed tests.
Projective tests In Chapter 2 the problems with projective tests, their

lack of reliability and validity, were discussed and it had to be concluded
that they were not satisfactory for the scientific analysis of personality
although objective scoring schemes might be able to remedy some of the
deficiencies. As a result of these psychometric defects, the results from the
clinical application of projective tests do not enter into the psychometric
view of personality despite the fact that the Rorschach test has been
extensively used in this field. However, because projective personality tests
play so large a part in clinical psychology some mention of the results will
be found in a brief section of this chapter.
Objective tests These tests which, as was made clear in Chapter 2, are
still largely of unknown validity, are potentially powerful in clinical
psychology since, in principle at least, they are measuring meaningful
factors. However, in practice, with insufficient evidence of validity,
interpretation of clinical results is difficult.
On these grounds much of this section on the contribution of the

psychometric view of personality to clinical psychology will be concerned
with the relevant findings from personality inventories, especially those
measuring clear factors.
Clinical diagnosis
Although diagnosis in clinical psychology is regarded by some psychologists
as anathema, being an example of an outmoded medical model and of
labelling clients to their disadvantage, in fact diagnosis is essential for the
scientific study of psycholobrical disorders. Only by accurate diagnosis can
psychologists feel confident that they are talking about the same condition,
when they are investigating causal factors and the efficacy of treatments.
Factored penonality tests in clinical psychology As has been dis-

cussed throughout this book, factor analysis has identified the most
important factors in the personality sphere. The big five factors, extraversion,
anxiety, tough-mindedness, conscientiousness and open-mindedness, can
usually be found in questionnaires among normal subjects and Cattell (e.g.
1973) isolated a number ofabnormal factors which are not so well replicated.
However, because the emergence of the big five factors is relatively recent,
there is little fundamental clinical work with these factors. Indeed two
psychologists stand out in their attempts to utilise the findings from the factor
analysis of personality tests in clinical psychology - Eysenck and Cattell -
and consequendy much of the work discussed below concerns their factors.
The work of Cattell

In principle, as Cattell and colleagues have argued (Bolton, 1986), there
are two approaches to clinical diagnosis using factored tests. These have
been discussed in the previous section on the use of tests in selection and
will be briefly mentioned here.
The regression method Here the multiple regression is computed

between the diagnostic group and the personality test scores. The weights
from the regression equation are used in the computation of the client's
score. The highest score represents the diagnostic category.
The matching method Here the profile of scores on the personality

test of various clinical groups is matched to the profile of scores of the
client. The best match represents the diagnosis. Cattell has developed a
special pattern similarity coefficient to ease the computation.
As was also mentioned in the previous chapter, the efficacy of both these
methods depends on adequate sampling of the diagnostic groups. In
addition, in the regression method it is essential that a satisfactory multiple
correlation was obtained and that the beta weights were replicated. In the
matching method it is important that there are clear differences between
the profiles of the groups. If all these conditions are fulfilled then these
methods are powerful. What they both reveal are the personality factors
which discriminate the diagnostic clinical groups.
Advantages of these methods The great advantage of using these

methods with factored personality tests is that because the psychological
meaning of th~e factors and their importance in personality is established,
all results are themselves meaningful. Thus if it turns out that extraversion
discriminates a particular clinical group then this is important in under-

standing the nature of the disorder, and similarly with other factors.
In fact Cattell and his colleagues have collected together a considerable
body of findings with the 16PF normal factors and the abnormal factors of
the CAQ (Krug, 1980), as well as with the MAT (Cattell, Horn and
Sweney, 1970), a motivation test and the objective OAB (Cattell and
Schuerger, 1976).
As has been pointed out in previous chapters of this book, the validity
of the Cattell personality tests, both of temperament and motivation, has
been called into question so that the substantive findings from this work
must remain questionable. Thus the main contribution of Cattell to this
field must be one of principle and method. There can be no doubt that his
approach of using factored variables to discriminate diagnostic and clinical
groups is potentially powerful once the major factors of personality have
been identified. On account of his brilliant and pioneering efforts in the
clinical field his main findings deserve a brief summary. For greater detail
readers must be referred to Bolton (1986).
According to Bolton (1986) it can be argued from the work with the
16PF test that diagnosed neurosis (incidentally a very broad category) is
caused by a combination of six primary source traits: low ego strength (C),
submissiveness (low E), desurgency (low F), emotional sensitivity (I), guilt
(0) and excessive ergic tension (Q4). It should not escape readers that this
factor analytic diagnosis of neurosis resembles to a remarkable degree the
basic psychoanalytic claims that neurosis results from the repression of a
too powerful superego. The high Q4 and 0 and the low C fit peculiarly
well. This agreement between the entirely speculative approach of psy-
choanalysis and the rigorous statistics of factor analysis is indeed striking.
In addition using the second-order factors it can be shown that neurotics
are, compared with normals, high on the anxiety factor and low on
extraversion.
Another important finding from the application of factor analytic
personality tests in the abnormal field supports the claim that in most
instances, psychopathology can be seen as an extreme of a normal con-
tinuum. Neurotics in general have personalities that are not qualitatively
different from those of normals.
This is particularly interesting because in the case of the psychoses
(abnormal syndromes characterised by lack of contact with reality and best
exemplified by the schizophrenias) this is not the case. Here the discrimi-
nating factors are not found in normal personality and are those factors
measured by the CAQ (Krug, 1980) which was described in Chapter 5.
Thus psychotics are qualitatively different from normals. For example,
schizophrenics are characterised by high scores on depressed withdrawal

and on the schizophrenic scale itself which deals with extreme withdrawal
from reality.
Again it is interesting to note that this distinction between neurotics and
psychotics fits well with psychoanalytic theory in which neurosis is seen as
the result of ego defences which all normals use while psychosis occurs
when the ego defences are overwhelmed.
At this point, mention should be made of depth psychometry which
has been well described by Heather Cattell (1986). The principle of depth
psychometry claims that the same general neurotic symptoms may result
from different configurations of the primary source traits which are
implicated in neurosis. This configuration is important because, as has been
argued, source traits, as revealed by factor analysis, are the underlying
determinants of the observed variations in human personality. These, in
the case of abnormal psychology, are the descriptive syndromes such as
anxiety neurosis or depression. Thus depth psychometry, by analysing the
underlying source traits, enables the psychologist to understand rather than
simply describe the diagnostic categories.
Furthermore, the fact that normal factors are implicated in neurotic
syndromes and abnormal factors in psychotic disorders immediately sug-
gests the importance of these factors for the assessment of therapeutic
progress. Thus in the case of neurotics, as treatment proceeds the elevated
scores on the normal factors should gradually approach the normal band.
Similarly in the case of psychotics, scores on the abnormal factors should
approach zero, since these factors are rarely seen in normals.
Bolton (1986) reports similarly powerful discriminations of neurotic
disorders with the T factors from an objective personality test - the
Objective Analytic Battery, the OAB (Cattell and Schuerger, 1976).
However, I shall net discuss these findings here because the validity of this
test is not well supported partly because it has not been widely used. Indeed,
a study of the validity of this test by Kline and Cooper (1984b) was not at
all encouraging, as was discussed in Chapter 4. However, in principle, if
valid objective personality tests could be devised they would be valuable
in abnormal psychology, used as has been described above.
Cattell's motivational factors, as measured by the MAT (Cattell, Hom
and Sweney, 1970), a test described in Chapter 6, have also been employed .
in the study of abnormal groups. The MAT factors can be placed into
regression equations or matched to the profiles of clinical groups, just as
was the case with the temperamental factors. Indeed, all these tests form
part of the psychometric model together with ability factors. However, as
Bolton (1986) has argued, the motivational factors have been used in a
different way. Again my description will be briefsince there are difficulties

with the validity of the MAT, as was shown by Cooper and Kline (1982).
The underlying dynamics of patients in the different diagnostic cat-
egories can be worked out. For this, P factors are computed for each
individual (involving the factoring of scores over occasions) and these are
then inserted into the dynamic calculus which indicates how drives are
expressed in behaviour, a quantified psychoanalysis according to Cattell
(1985). Although the mathematics of this dynamic calculus are exceedingly
complex, a simple illustration will clarify its essence. IfP analysis revealed
that in a particular individual the sex erg loaded on a factor which included
activities such as carrying out factor analyses and writing short stories it
would imply that these activities had sexual roots, as indeed, psychoanalysis
supposes. In this way the dynamics of behaviour can be revealed.
Since the dynamic calculus is highly abstruse and since it is not accepted
other than by Cattell and his colleagues and since there is doubt, as has
been argued, concerning the validity of these ergs and sentiments, no more
details of these methods will be given. Nevertheless despite these problems,
the principles of the dynamic calculus are brilliant and original and ifbetter
measures could be developed this approach would be extremely powerful.
There is one fmal use of factored tests in clinical psychology which
deserves brief mention. This concerns their application in the study of
therapeutic success. One example has already been mentioned, namely that
as treatment proceeds, we should expect the elevated scores on the salient
factors to return to the normal level. This is useful and important.
However, there is a far more sophisticated approach to the problem of
therapeutic recovery.
This involves putting into factor analyses measures of temperament and
dynamics together with details of what went on in the therapy, type of
therapist, current events in the life of the patient and therapeutic outcome.
From the variables loading on therapeutic outcome it would be possible
to answer questions which dog clinical psychology, such as the effect of
life events, type of therapist and therapeutic interventions. Unfortunately,
owing to the dislike of quantification among the majority of clinical
psychologists, no such studies have as yet been carried out, although case
work with the Cattell factors (Heather Cattell, 1986) strongly supports this
approach. Since it is possible that some of the links in these processes are
not linear but involve interactions, more complex analyses such as are
found in structural modelling Qoreskog and Sorbom, 1979) might be
valuable.
Conclusions This discussion of the use by Cattell and colleagues of
Cattell's factored tests indicates clearly the brilliance of Cattell's work both
in relation to diagnosis and to treatment. These methods could lead, as was
suggested, to a quantified psychoanalysis. However, the fact that the tests
are probably not valid, except for the second-order personality inventory
factors, means that substantive findings are to be treated with caution.
However, it should be said that as better tests are developed they could be
used with the methods which have been discussed in the development of
an effective psychometric clinical psychology.
The work ofEysenck

Eysenck and his colleagues at the Maudsley have carried out considerable
clinical research with their factors, E, Nand P, as measured by the EPQ.
These of course, as has been argued, have extensive evidence for validity,
are second-order factors and are three of the big five factors which regularly
occur in personality questionnaires. It should also be noted that the first
two of these factors are identical with the first two second-order Cattell
factors- exvia and anxiety.
In respect of E and N the findings are highly similar to those of Cattell
with these second-order factors. Neurotics tend to be unstable introverts.
Of course in the Eysenck system N is regarded as arising from lability of
the autonomic nervous system (Eysenck, 1967), while extraversion is
conceived as stemming from the arousability of the central nervous system,
and it is alleged by Eysenck (1976) that extraverts are more difficult to
condition than introverts, and that the differences between abnormal
groups on these two factors are accounted for by these physiological factors.
The P, psychoticism, or tough-mindedness factor is particularly inter-
esting. As Eysenck and Eysenck (1976) demonstrated, P discriminates as
surely as does N, psychiatric groups from normals. Psychotic patients (made
up mostly of schizophrenics) and prisoners have the highest scores, fol-
lowed by drug addicts, patients with personality disorders and sex offenders.
Furthermore, although the P scale is not a symptom list, severity of
symptoms is positively correlated with P.
In discussing these findings the power of using meaningful factored
scales, rather than criterion-keyed tests of unknown psychological
meaning, becomes apparent. Thus, for example, theN factor is also good
at discriminating clinical groups, in this case mainly neurotic groups.
However, P and N are orthogonal, uncorrelated. From this discussion the
following conclusions may be drawn.
Neuroses and psychoses are not on the same continuum. Thus neurotics
score more highly than normals or, N, psychotics on P. This means that
psychotic and neurotic disorders are qualitatively different. Psychosis is not

simply an exaggerated neurosis. Actually Claridge (1985) has developed an
excellent psychometric measure of schizotypic personality with which he
has shown that there are certain schizophrenic-like symptoms possessed by
normals which are exaggerated in both schizophenics and gifted, creative
individuals. On this psychometric basis he has developed an impressive
theoretical account of the development of schizophrenia. His work is an
excellent example ofthe value ofpsychometrics in clinical psychology even
though it is likely that the schizotypal personality is not factorially clear.
Since, as was discussed in Chapter 7, all these factors have a considerable
genetic determination, all these results are of considerable theoretical
interest for clinical psychology. It certainly appears that these three per-
sonality factors are highly important in the development of neurotic and
psychotic disorders and hence in their treatment. Any adequate clinical
theories must take them into account. It must be concluded even from this
brief discussion of the contribution ofEysenck to clinical diagnosis that the
psychometric view of personality is of !,'Teat psycholo~:,>ical significance.
Findings from other personality tests

In a book in which clinical psychology is only a small section it would not
be possible to discuss all the results of the administration of personality tests
in the clinical field. However, as has been made clear, the main contribu-
tion comes from factored tests which by definition deal with the most
important variables in the field. Nevertheless certain non-factored person-
ality tests have been so widely used in clinical psychology that their results
must be examined.
The MMPI and MMPI-2

These tests were described in Chapter 6 and in Chapter 2 where it was
pointed out that despite the fact that the original test, the MMPI, was the
most widely used personality inventory, the fact that it was criterion-keyed
rendered the scales empty of psycholo~:,>ical meaning. Thus the mass of
clinical findings reported with this test have made and can make little
contribution to knowledge of clinical psychology. All that can be said is
that an MMPl scale (and more than 200 have been developed from the
original item pool, Dahlstrom and Walsh, 1960) can or cannot discriminate
particular criterion groups. Unfortunately the same is true of the newly
developed MMPI-2 (Graham, 1990).
Proponents of the MMPI can argue that for some purposes discrimina-
tion is all that is required. Thus if perhaps in army selection it is necessary

to screen out all neurotics or psychotics, an efficient criterion-keyed test
such as the MMPI is highly useful. While this is true, it should be pointed
out that there is no reason why a battery of factored tests should not be
equally efficient at screening out clinical cases and in this case some
understanding of the psychological nature of the problems would have
been gained.
Since, in addition, the diagnosis of clinical groups is a problem in itself
and is highly unreliable (Beck, 1962) the whole basis of the original keying
of the MMPI items is a difficulty. This together with the fact that the scales
are not reliable and have overlapping items also renders the psychometric
qualities of the MMPI dubious.
Since the test is so widely used in clinical psychology, attempts have
been made to factor it and thus render it a more useful test, as was discussed
in Chapter 5. At the scale level there is general agreement that two factors
account for much of the variance - anxiety and ego strength (Friedman et
al., 1989) -factors which come as no surprise given the fact that these are
implicated in the clinical work with the 16PF test, discussed above.
Rather than scales, parcels of items were factored by Cattell and
colleagues and some of these factors were used in the CAQ (Krug, 1980)
which attempts to measure the main abnormal factors and which has been
described in this chapter. Probably the ability of the MMPI to discriminate
clinical groups depends on these factors.
Johnson et al. (1984) factored the MMPI items using 11,000 subjects.
They obtained a large number of factors which generally grouped together
the items with similar meanings. However, they were identified simply
from their item content with no external validity and thus this identification
must remain uncertain. However, they identified anxiety, psychoticism
and extraversion, all part of the big five and paranoia, psychopathic
deviation and psychasthenia which are included in the CAQ (Krug, 1980).
Costa et al. (1985) also claim to have found the big five, as they do in every
test, in the MMPI items. Finally it should be noted that Graham (1990)
argues that the Johnson et al. (1984) factors are probably to be found in the
MMPI-2.
Conclusions Despite the wide use of the MMPI in clinical studies and
the updating of the MMPI-2 it is difficult to justify the work. It is only
suitable for screening and the lack ofmeaning of the scales makes it inferior
to factored tests. The clinical contribution of psychometrics cannot arise
from either of these tests or from a!ly criterion-keyed tests. That it measures
factors is fortuitous and results from the skilful choice of items in the
original item pool. However, these factors are better measured by tests
deliberately designed to do so. The best of the MMPI is to be found in the
second part of the CAQ. In brief, a test that has not contributed to the
psychometric view of personality in abnormal psychology.
The Rorschach Test

This projective test (Rorschach, 1921) is one of the most famous person-
ality tests known beyond the confines ofpsychology. Rorschach developed
this series of ten inkblots specifically for clinical work and over the years
there have been thousands of clinical studies and some different versions
of the test have been constructed.
However, as was discussed in Chapter 2 the Rorschach is not reliable
and has poor evidence for validity. However, G analysis of Rorschach data
(Holley, 1973) where objectively scored Rorschach protocols are subjected
to multivariate analysis might prove useful in the clinical context. So far
little of substantive value has been attained by these methods but they are
worthy of further research because they make use of the rich data which
are yielded by projective tests. Furthennore these projective data, as has
been argued in Chapter 2, may be valuable since some researchers appear
to be able to use them effectively. It should also be pointed out that Exner
(1986) has attempted to put Rorschach scoring on a more objective
footing, while keeping within the boundaries of normal projective test
scoring. Even so there is not strong evidence for the validity ofRorschach
interpretations.
In brief, at present it cannot be argued that the Rorschach has made a
powerful contribution to the psycholob'Y of personality either normal or
abnormal and it is unlikely to do so until reliable quantification is intro-
duced.
Conclusions From this discussion of two of the most famous non-fac-

tored tests commonly used in clinical psycholob'Y· it is clear that little
psychological knowledge has come from them. The advantages of using
factored tests with externally validated psychologically meaningful factors
become obvious when the problems of the MMPI and Rorschach are
discussed.
Contributions to clinical theory

There are two contributions to clinical psychology from psychometrics.
The first and less important is as a test of other clinical theories. Thus if, as
is the case with ergs and sentiments discovered by factor analysis (Cattell,
1985), it appears that there are ten important motivational factors, then
clinical theories which deal with motivation and which posit a different
number of variables are shown to be incorrect. To continue with this
example Freudian theory has two drives, sex and aggression, or later Eros
and Thanatos. These are underestimates. McDougall (1932) and Murray
(1938) considerably overestimate the numbers of drives. Again, writers
such as McClelland (1961) who stress the importance of need achievement
have probably made too much of this variable.
A second and potentially more valuable use of the psychometric factor
analytic findings in personality is to construct clinical theories on the basis
of the actual results. The rationale for this is that the simple structure factors
are, as has been argued, the salient variables in the field. This being the case
any adequate theory must account for them. In clinical psychology it has
been shown, must for example, that Eysenck's extraversion, neuroticism
and psychoticism are salient variables in the discrimination of neurotics and
psychotics and any theory ofabnormal psychology must, therefore, include
these factors as critical variables.
Indeed, the specification equations of personality factors for clinical
diagnostic groups, if the multiple correlations are high, effectively con-
stitute theories. Thus if three personality factors can predict that an
individual is a depressive, then these are all that is necessary to understand
the condition.
So far in clinical psychology, owing to the problems ofisolating the best
set of personality factors, substantive findings from the specification equa-
tions have not yet been made, although the attempts to do this by Cattell
are models of how a genuine clinical psychometric theory could be
constructed. Actually, in the next chapter, we shall see how Eysenck and
Cattell have attempted to develop more general psychometric theories of
personality based upon their factor analytic results.
Educational psychology
Educational psychology is concerned with the psychological problems
affecting educational progress and achievement. Thus to some extent there
is bound to be some overlap with clinical psychology, if emotional
difficulties are at the root of educational problems, and with occupational
psychology when questions of selection and vocational guidance arise.
In this section of this chapter I shall briefly set out what personality tests
have to offer to various problems central to educational psychology as it
has been defined ·above.
The failing child Educational psychologists are frequently asked to

investigate the causes of failure in children, when their teachers are at a loss
to account for the poor performance. I do not want here to make
exaggerated claims as to the importance of personality tests in this aspect
of educational psychology. As is fully documented in any modern text of
educational psychology (e.g. Fontana, 1981) there are various factors which
can affect a child's performance at school: attitude of parents to school;
attitude of peers to school; having somewhere quiet to work; emotional
problems of the child; level of intellectual ability; quality of the school, just
for example.
Given these complex and doubtless interacting factors at work in
determining academic performance, it may well be doubted whether there
is any useful role for personality testing. In fact there is a small but limited
place for personality testing, although this is restricted to factored person-
ality tests because only these have the necessary reliability and validity.
Projective and objective tests are not satisfactory except as research instru-
ments in this field.
The rationale for the value of personality tests in this aspect of educa-
tional psychology is to be found in the extensive work that was carried out
in the 1960s and 1970s in predicting academic success, much of which is
summarised in Kline (1979; Cattell and Kline, 1977). In almost all cases
the tests used were those of Cattell and Eysenck, which, as has been
discussed, are among the best factored measures. Here it was found that at
the primary school, mildly neurotic extraverts did best, while at the
secondary level the anxious introvert was successful. At university the
findings are essentially similar to secondary school, not only in the West
but in cultures such as India and Africa, for example Ghana (Kline, 1966).
Particularly important in these studies is the work of Cattell and Butcher
(1968) who put into a multiple regression with academic success at the
secondary school, tests of ability, personality and motivation, thus allowing
the relative predictive power of these variables to be assessed. The critical
points to note about these results are that the correlations with personality
and motivational factors, while significant, are small, in the region of0.2
to 0.3, and that the contribution of the three types of factor - ability,
temperamental and dynamic - is separate and distinct. Thus Cattell and
Butcher (1968) found a multiple correlation of0.6 with academic success
which is about as high as it could be if there are any effects at all of school
and environment on academic performance.
Given the small size of the individual correlations between success and
dynamic and temperamental factors, interpretations of the scores of any
individual would have to be extremely cautious. Furthermore the person-
ality tests would have to be interpreted in the light of the results of ability
tests. Thus if a child was simply not bright enough for a highly academic
school, this is sufficient reason for difficulties. If, however, failure is not
attributable to this or any other important family or domestic factors, as far
as can be ascertained, then it makes sense to examine the personality and
motivational factors which are correlated to academic success. For
example, if it turns out that a child is highly extravert, that might at least
in part be a contributory cause of difficulty. Here, it might be helpful to
put the child in as extraverted an atmosphere as was possible at the school,
for example with the most lively and noisy teachers.
Sometimes personality testing can reveal that a child is highly anxious.
Although it is difficult to remedy this, teachers can bear this in mind in the
way they treat the child and this might produce some improvement. As a
final example, it could be that a child has a low N score. Now this, as has
been seen in the correlation with anxiety, is not conducive to high
academic performance, presumably because if students are low on anxiety
they do not worry ifassignments are not completed nor are they galvanised
into study by the imminence of examinations. These small correlations
with anxiety do not suggest, incidentally, that high anxiety is good for
academic success. It is almost certainly not. Rather there is a moderate
facilitating level which tends to make students work hard, worry about
failure and seek to remedy their failings - all leading to enhanced examin-
ation performance.
From this discussion it is clear that there is a part to play for factored
personality tests in the diagnosis and treatment of educational difficulties
although all test scores must be interpreted in the light of the abilities of
the students and their other personal circumstances.
Educational selection The fact that there are significant correlations

between academic success and Cattell's temperamental and dynamic factors
(which do not overlap with ability factors) indicates clearly that in any
rational selection system for education these personality factors would have
to be included.
In the handbook to the 16PF test, regression equations for academic
success at university can be found and these would be used exactly as was
suggested for the regression equations in occupational selection. However,
it is clear that more work is required in this field since the four regressions
in the text have somewhat different weights, although the multiple
correlations are substantial. These fluctuations are probably related to the
unreliability of the Cattell factors, as has been discussed.
The fact that the individual correlations with the primary factors are
small and that the weights vary from study to study means that in practice
it is better to include the second-order factors or the big five factors which
have been shown to embrace much of the reliable personality variance. If
this were done it would probably eliminate those whose personality was
unsuited to academic pursuits- extreme extraverts, the highly anxious and
the high P scorers. Those who were high on conscientiousness would be
favoured by such a system and this is correct since, in our present
educational system, these individuals do well relative to their abilities.
In an egalitarian age such as ours, the notion of including personality
tests for selection to education is undoubtedly impractical, since there is
resistance even to the application ofiQ tests when there can be no doubt
that educability depends to a considerable extent on intelligence Qensen,
1980). Nevertheless, given the correlations, if personality tests were used
in selection procedures together, of course, with tests of ability, overall
selection would be more efficient.
Vocational guidance I shall not say much here about the use of
personality tests in vocational guidance because the basis of so doing is
identical with that in occupational selection which has been discussed in
detail at the beginning of the chapter. Exactly as was there argued, the
psychometric model underpinning the use ofall psychological tests suggests
that there is an ideal configuration of personality traits for each job. These,
therefore, have to be discovered and then each individual can be tested to
find the job that is most appropriate.
Thus personality testing is an essential part of vocational guidance. It is
not the place in this chapter to discuss in detail how the personality test
scores should be used for vocational guidance (see Kline, 1975, 1992a).
Suffice it to say that they would fonn the basis for discussion with the
children rather than being used in a rib>id statistical formulation, as some-
times occurs in selection. However, there is no doubt that those factored
personality and motivation tests which correlate highly with occupational
success or discriminate clearly different occupational groups are critical
variables to be discussed in any efficient system of vocational guidance.
Educational theory
I shall now discuss the contribution from the psychometric analysis of
personality to educational theory. Gillis (1986) attempts this task in relation
to the Cattell personality factors but his arguments are applicable in
principle to other factors . He takes the ret,>Tession equations of the 16PF
tests for academic success (discussed in the section on selection) and those
of the HSPQ (the high-school version of the 16PF) where it appears that
superego and self-sentiment are the best predictors. These are, of course,
dynamic variables as well and when measured as such in the MAT they
also have the highest weights.
Clearly, as Gillis (1986) argues, these findings should be incorporated
into any theory which attempts to account for educational success. In fact,
it fits the work of McDougall (1932) who regarded superego and self-sen-
timent as master sentiments, the keys to understanding human dynamics.
While this is so, it must be pointed out that such findings cannot be used
to prove this theory since proof requires refutation not confirmation
(Popper, 1959). Of course, factored results can be used to refute any theory
or to provide support, as this example illustrates.
In addition they can be used to develop a psychometric theory. Indeed,
both Eysenck (1967) and Cattell (1981) have attempted to fit their findings
into a theoretical framework. However, this is broad and general rather
than specific to educational psychology and a.s such it will be discussed in
the next chapter.
Conclusions
In thi~ section the salience of personality testing in educational psychology
has been made clear. Meaningful, factored personality tests have a small
part to play in the study of individual problems in educational success and
can undoubtedly improve the efficiency of educational selection. They are
essential for vocational ~idance and can be useful in the development of
educational theory.
Chapter 9
Personality theory and the

psychometrics of personality
As has been argued throughout this book, the psychometric view of

personality is based upon the results of psychometric personality tests, not
all tests however, but factor analytic tests, where the factors have been
externally validated and are not simply collections of semantically similar
items. Such factors are, from the nature of factor analysis itself, the most
important variables in the field because they account for the maximum
variance. Thus if factor analysis has been properly executed - adequate
sampling of tests and variables, the selection of the correct number of
factors, rotation to simple structure (and possibly further confirmatory
analysis) followed by external identification of the factors - it must be the
case that resulting factors are the salient variables and, therefore, the ones
that must be central in any theory of personality.
With this view offactors it is clear that the factor analysis of personality
bears upon personality theory in two ways. On the one hand it is highly
pertinent to all theories of personality, of which, as was pointed out in the
introductory chapter, there is a very large number. Any theory which does
not deal with the main factors must be missing the point. On the other
hand, a psychometric personality theory can be developed utilising the
main factors from the analyses.
In this chapter on personality theory both these aspects of the work will
be considered, and I shall deal first with the bearing of the results on
personality theories. It is not possible or even desirable to examine the
bearing of the factor analyses of personality on all personality theories, since
many theories of personality, especially those developed early in this
century, are little more than interesting speculations (and often they are
not even interesting) . However, I shall consider the most famous and
influential of these theories, which are not obviously shown to be false by
other current developments in psycholob'Y, together with some of the more
132 Personality theory
recent accounts which are popular in modem texts on personality (e.g.

Pervin, 1990).
The scientific status of factors

Before the bearing of the findings from the factor analysis of personality
on personality theories can be evaluated it is necessary to scrutinise the
status of factors a little more carefully. So far in this book the strictly
statistical view has been argued that factors are a set of dimensions which
account for the intercorrelations among the original variables. As Cattell
(1978) has argued in cases where the factors are known, from theories of
physics for example, simple structure factor analyses are able to find these
factors, and they can be regarded as causal detenninants. Nevertheless, the
status of personality factors in a field where the notion of determiners is
not so clear cut, requires further examination.
To clarify the points 1 shall deal with one factor, extraversion, rather
than the whole set, although, of course, the arguments apply equally to all
personality factors.
Extraversion can be thought of as a construct which accounts for a good
proportion of variance in personality questionnaires. More specifically, it
accounts for the intercorrelations between certain kinds of items: those
concerned with enjoying parties, being sociable, liking people and excite-
ment and so on. In this sense extraversion might be said to have a causal
function.
However, what does it mean to say that extraversion is a personality
factor? It is clear that extraversion is not a physical object such as a brain
or an eye. At no point will neuroscientists be able to open the skull and
point to extraversion. Extraversion is a concept, or construct, defined not
verbally but by its factor loadings. It is a concept, therefore, of higher
precision than one simply verbally defined. In science, constructs are no
less important than physical objects which, of course, are also constructs.
With no concept of brain for example, it could not be separated from the
head or regarded as a single entity and not a collection of neurons. An
example of a highly useful concept in biology is that of Man. This includes
all groups of Homo sapiens, regardless of skin colour or height for example,
variables in which there are considerable variations, and other now extinct
forms. Now there is no object Man which could ever be observed,
although the concept is essential in understanding the development of
many human characteristics. It should be pointed out at this point that if a
large number of biological characteristics of mammals were subjected to
factor analysis, the major Linnaean b'l'oups; primates, carnivores, and -
Personality theory 133
within them, dogs and cats for example - would undoubtedly emerge as
factors. Just as these factors account for mammalian physical characteristics,
so extraversion is essential in understanding the fact that certain traits are
found together, as well as other important aspects of personality as will be
discussed later in this section, especially when the work of Eysenck is
described. In other words, the fact that extraversion is a concept docs not
mean that it is scientifically valueless. This is particularly so since personality
itself is an abstract concept. Indeed most of psychology, especially the study
of personality, is a field of abstract concepts rather than physical objects.
The physical in psychology is the province of anatomy and neurology and
it is arguable that this primarily conceptual nature of the subject makes it
unsuitable to the methods of the natural sciences. This indeed was the view
of the behaviourists who attempted to remedy the fault by studying only
observable behaviour. Nevertheless, since science does deal with abstrac-
tions, this is probably too pessimistic a view.
There are several advantages of factor analytically defined concepts
compared with verbal notions. One mentioned above is precision. Factor
loadings are numerical and the tests are specified. Verbal definitions are
notoriously slippery. The second and more significant advantage, is rooted
in the nature of factor analysis. Factors are important because they account
for variance. Now it is possible to develop concepts by speculation or
philosophy which empirically turn out to be trivial or misleading. To take
again some obvious biological examples, it would be possible to classifY
mammals (if that was a recognisable category) by fur colour of by density
of fur. The length and presence of tails might be another classificatory
factor. These, as it turns out, are entirely superficial and trivial charac-
teristics in understanding different mammalian fonns. Now it could well
be the case, since the psycholob'Y of personality is still in an elementary
state, for there is little agreement in the field, that many of the variables
and concepts in usc are similarly superficial and trivial.
The main personality factors

As was made dear in Chapter 4, five personality factors appear to account
for the variance in most personality questionnaires. Of these the three most
important, and the backbone ofEysenck's system, have the highest validity
of any personality measures and their external correlates are particularly
well supported. These factors are: extraversion, neuroticism and psychotic-
ism or tough-mindedness, factors at the second-order to be found in the
system of Cattell. The other two of the big five factors are open-minded-
ness and conscientiousness. Clearly, as has been argued, any viable person-
ality theory must incorporate these factors as important variables.
As was made clear, the dynamic factors revealed by factor analysis are
not so well supported as the temperamental factors. Only Cattell (1985)
has carried out systematic factor analytic research into personality dyna-
mics. Despite the relative uncertainty of the list of ergs and sentiments,
these results are still useful in assessing the worth of dynamic theories of
personality, although their results have to be treated with more caution
than is the case with the temperamental factors. Cattell's work on the
strength of interest factors should also not be forgotten although this is even
more problematic than the work on the structure of drives.
Freudian theory Whatever the virtues of Freudian personality theory,

as found in Freud (1933), among them·cannot be found the explication of
these factors . For psychoanalysis, overt personality is to be seen as resulting
from the efforts of the ego, especially defence mechanisms, to hold the
dynamic equilibrium between ego, superego and id. In this conception of
personality there is no place for these three factors.
Of course, Freudian psychoanalysis is a complex theory. Another aspect
of Freudian personality theory is developmental. Thus fixation at certain
levels of development, psychosexual phases (Freud, 1905), was said to lead
to the development of oral, anal or phallic characters.
Kline (1981) has studied in detail the psychometric evidence concerning
these Freudian personality syndromes and has developed and validated
measures of the anal and oral character (Kline, 1971; Kline and Storey,
1977). If the putative psychosexual origins of these personality constella-
tions (hence their names) are ib'l1ored, there still remains the question as to
whether such syndromes can be observed. It could be the case that Freud
had correctly observed the personality characteristics but quite wrongly
inferred their developmental bases.
In fact it appears that there are two personality syndromes, oral pessim-
ism and oral optimism, that correspond well to the oral character and one
that fits closely the anal character, although there is little evidence to relate
these syndromes to infantile experience, other than a study by Kline (1968)
in which the scores on Ai3Q, ameasure of anal traits, correlated positively
to a quite different measure of anality - that of the Blacky Pictures (Blum,
1949) which portrays a defecating dog. This was an interesting finding
rather than compelling evidence for the link. Factor analytic studies of these
syndromes showed that, despite the outlandish names, there was no little
resemblance to the main psychometric factors described in this book. Thus,
the oral optimistic constellation correlated significantly and positively with
extraversion. The oral optimist is a stable extravert. The oral pessimist on

the other hand is anxious. Thus these two syndromes essentially describe
the extraverted and the anxious individual. The relationship of the anal
character to the big five was even more striking. In a study of the
authoritarian personality (Kline and Cooper, 1984a) in which Ai3Q was
factored together with the 16PF test and a number of other measures of
authoritarian personality, a clear obsessional or authoritarian factor
emerged. As was argued in Kline (1992a) this factor resembles the openness
to experience and conscientiousness factors of the big five. Thus in the
concept of the anal character, there seems little doubt that Freud described
a mixture of two of these big five factors.
At this point in the argument a brief aside is in order concerning the
authoritarian personality. It does appear, as has been shown in earlier
chapters of this book, that it is an important personality syndrome,
characterised by obeisance to superiors and tough-minded control of
inferiors. Genetic studies ofthis variable could be valuable since it is possible
that this factor can be observed in the hierarchical ordering of many
mammalian groups. This might account for the prevalence of authoritarian
attitudes in human societies and the difficulty of eliminating them despite
their unpleasant consequences.
To return to Freudian theory descriptively, Freud's psychosexual per-
sonality theory does not seem wide of the mark, although as has been seen
these syndromes are mixtures of factors and the claimed aetiology is
woefully short of empirical evidence. However, psychosexual personality
syndromes are a relatively minor aspect of Freud's notions of personality
and the factor analytic findinb"S in the field of temperament are not really
germane to his theorising, with its stress on the unconscious determinants
ofbehaviour.
As regards the pertinence of the dynamic factors, ergs and sentiments,
to Freudian theory, it must be said that it is not supported by these findings.
Thus if psychoanalysis had been correct, two factors would have been
expected, sexuality and a~>Tession (even Eros and Thanatos) or possibly,
only one- sexuality. However, this is not the case and there appears to be
a longer list of dynamic factors.
However, the factor analyses of strength of interests which resulted in
three main factors: alpha, conscious id; beta, realised ego; and gamma, the
conscience factor, suggest that strength of interest is determined by three
components of which only one, beta, is directly rational. This provides
some confirmation for the Freudian notion that interests are determined
by both conscious and unconscious factors. These results support the
Freudian views in principle rather than in detail. They also cast doubt on
the face-valid simple inventory approach to the study of attitudes and

interests, a method which is concerned solely with the beta factor.
Other psychoanalytic theories

I do not want to discuss in any detail the pertinence of these factor analytic
findings to other psychoanalytic theories because their influence in modem
psychology is small. However, there are two points of interest.
First, Jung (1949) developed a typology of personality in which
extraversion was an important category. Individuals were either extraverts
or introverts and there were four categories of each. However, although
extraversion is one of the main personality factors in the factor analytic
conception, extraversion is a continuum and individuals may be placed
anywhere on it. Jung's notion of an extraverted type is not supported by
the evidence concerning the distributions ofscores on tests, although these
are of course dependent on the particular sets of items. Furthermore there
has been no clear support for the eight fold categorisation, despite the
popularity of the Myers-Briggs Type Indicator (Brig{,"S and Myers, 1962).
Second, Adler (e.g. 1927) claimed that the basic drive for man was the
upward striving for superiority to overcome feelings of inferiority and that
most human behaviour could be understood in this way. This of course
implies that there should be one large general motivational factor which is
not the case. Adler however did stress the importance of within family
differences affecting personality, notably family position, and this is sup-
ported by the genetic studies of the factors, discussed in Chapter 7.
Some other more recent theorists such as McClelland (1961) have
argued a similar case regarding need achievement as a paramount drive.
Again this is not supported by the factor analytic evidence which sees this
variable as a composite of factors, self-assertion, career and self-sentiment
being the most salient.
Conclusions In general these factorial fmdings give little support for

psychoanalytic theories of personality which, in the main, do not deal with
factors which have emerged as the most important, accounting for the most
vanance.
The work of Murray (1938) and McDougall (1932)

These are two of the most influential older theorists of personality and
motivation and the work of Murray continues to spawn new tests (see those
ofJackson, discussed in Chapter 5). McDougall (1932) aq,'Ued that propen-
sities were the basic drives, common to both mammals and human beings.
However, most human activities are driven indirectly by sentiments. As is
clear from this account, Cattell's motivational system is highly similar, in
that he sees activities as resulting from the action of ergs and sentiments.
Cattell, indeed, attempted to give McDougall's work an empirical, factor
analytic basis.
Among McDougall's propensities were: food-seeking, disgust, sex, fear,
protection, gregariousness, self-assertion, submission, anger, appeal, con-
structiveness, acquisition, laughter, comfort, sleep, migration, plus a
number ofbodily needs. From this list it appears that there is overlap with
the ergs of Cattell, although McDougall claims more ergs than Cattell has
replicated. Although there is some agreement with the hypothesised ergs,
it must be remembered that Cattell's approach was influenced by McDou-
gall's work.
As regards sentiments, there is no doubt that here McDougall proposed
a far greater number than has been isolated by Cattell but his emphasis on
the importance of the self-sentiment does receive support from factor
analysis. In brief, there is surprisingly good s4pport for this work from factor
analysis although McDougall probably considerably overestimated the
numbers of drives and sentiments.
As regards the work of Murray (1938), factor analysis is not supportive.
He postulated on the basis ofhis personological studies which involved the
most intensive research of individuals as whole people - hence the name
- a long list of needs, few of which have been isolated by factor analysis.
This is probably because Murray's needs were surface traits, close to
observation, rather than the more fundamental source traits of factor
analysis. Furthermore the tests which are based upon his needs, of which
the best example is the PRF by Jackson (1974), technically an excellent
test, have not been validated as a measure of needs. Indeed, as has been
argued, there is evidence that the PRF is rather a measure of the big five
(Costa and McCrae, 1988).
Situationalism
Mischel (1968) originally argued that the measurement of traits was not
useful because in general correlations between traits and other criteria were
low, around 0.3, on account of the fact that there was less consistency in
behaviour than was assumed by the trait model, of which the psychometric
model is a precise form. This was because, it was argued, there was so much
variation in behaviour from situation to situation. This ofcourse is perfectly
true. Thus people shout at football matches and are quiet at funerals. In
these original formulations, Mischel had used the fact of situational

variation to argue that the notion of traits was not valuable and that trait
measurement, the core of the psychometric view, was essentially worthless.
In more recent publications (Mischel, 1984, for example), he has taken a
less extreme line and stressed that in understanding behaviour the situation
must be taken into account, but has admitted that personality traits are also
influential.
As was mentioned above, it is perfectly obvious that situations affect
behaviour. Furthermore, it is clear that situations interact with traits. Thus
individuals high on anxiety are far more likely to be upset, if trapped in a
cave, than those low on the variable, to take a simple example. However,
there are interactions which are less immediately obvious. Thus individuals
high on conscientiousness, one of the big five factors, react differently from
those low on the factor when put in situations where conscientiousness is
important - perhaps in vigilance tasks, such as examination marking. I do
not refer here simply to conscientiousness at the task, but at anxiety over
the accuracy of the scoring and the fairness of the procedure and so on.
Furthermore, it is apparent that over the long term there is real
consistency in behaviour. To take well-known political examples, Mrs
Thatcher was rarely submissive and Hitler was not ever tender-minded.
To those critics of these claims who argue that this is due to stereotyping
(Shweder, 1975), ask the relatives of the 6,000,000 who died in the
concentration camps. In addition the work of Block (1971) shows con-
siderable consistency in personality over time.
It should also be pointed out that many such political and social views
and values have a considerable genetic determination, as was shown in
some of the studies discussed in Chapter 7, notably Eaves et a/. (1989).
Finally, of course, if personality traits were of no importance in
understanding behaviour, the correlations with educational, occupational
and clinical criteria, discussed in the last chapter, and the considerable
hereditary, determination of the largest personality factors (see Chapter 7),
would be inexplicable. What is required in any adequate psychometric
view of personality is that situations be quantified and inserted into the
relevant specification equations along with the ability, personality and
motivational trait scores. This is a difficult project and one that psychometry
has only just begun, although Cattell (1981) has outlined some possible but
highly complex procedures.
Attribution theory
In our discussion of situationalism one of the aq,>Uments utilised by
situationalists to account for the apparent consistency of individuals in

everyday life was the notion of stereotyping, namely that people expect
certain individuals to behave in certain ways. It is these expectations which
create the consistency which is thus not real, in the sense of being part of
the individual, but is an attribution of the observer. How an individual is
cast, as it were, in her mould, depends on further stereotypes ofappearance,
national group, age, sex and the situations in which that individual is seen.
As Eiser (1980) argues, attribution theory treats trait descriptions as a
possible output from, rather than a possible input to, the process of person
perception.
There is no doubt that judgements concerning personality traits are
influenced by many of the cognitive and social factors discussed in
attribution theory. Furthermore it is perfectly reasonable for social psycho-
logists to elucidate the de~erminants of such judgements. It should be
noticed, however, that the personality traits of psychometrics are not social
judgements but factored test scores. Nevertheless that such social processes
occur does not affect the status of traits as important determinants of
personality.
Let me give an example from animal behaviour. It makes good sense
to describe a cheetah as speedy. The fact that people so judge it to be on
account of its muscular frame, its long legs and the reputation big cats have
for speed, is irrelevant as to whether or not a cheetah is fast. This is capable
of objective measurement. So it is with personality traits. Jews may be
judged as mean or Blacks as lazy, but objective measures of these traits can
ascertain whether such stereotypes are true or false.
As has been shown in our chapters on personality measurement, certain
personality traits can be measured highly reliably and with good validity as
judged by correlations with external criteria. Furthermore if traits were no
more than explanatory fictions, their correlations with occupational and
educational success and their high heritability ratios would make no sense.
Thus attribution theory, interesting as it may be to social psychology, does
not impugn the psychological importance of personality traits and thus the
It is quite clear, from the discussion in this chapter, that none of the
important theories of personality which have been scrutinised are sup-
ported by the psychometric results. In addition, there is one overarching
criticism which applies to all of them and, indeed, to almost all the
personality theories discussed in textbooks ofpersonality, as Eysenck (1990)
has argued. This concerns the results ofthe biometric analyses ofpersonality
where it was shown that between family differences did not contribute to
variations in personality test scores (only two studies running against this
trend). This is not only contra-intuitive but contrary to previous theories.
This means that it is necessary for psychometrics to develop its own
theoretical account of the findings. There have been two outstanding
attempts to do this by Eysenck (1967; Eysenck and Eysenck, 1976) and by
Cattell (Cattell and Kline, 1977; Cattell, 1981), and these will now be
described and evaluated. Both these writers have the most enormous
outputs- Eysenck has more than 1,000 publications and Cattell in the
region of500, many long and technically formidable book.,, The references
cited are the main sources of the theories but in a chapter of this length
complete citation is difficult.
'Tire work of Eysenck

To explicate the theory of Eysenck, I shall discuss the biological basis of
the personality factors as claimed by Eysenck (1967}, which was briefly
mentioned in Chapter 4, and then discuss how these are related to
conditioning and learning which are fundamental to understanding human
behaviour, at least in Eysenck's view. I shall also discuss briefly how these
theoretical views impinge on other important aspects of behaviour, such
as neurosis and crime. Indeed, one of the most impressive features of
Eysenckian personality theory is its applicability to real life as distinct from
the psychological laboratory, although it is claimed also to be salient there.
Physiological factors In Eysenckian theory, extraversion is related to

differences in cortical arousal (reticular formation activity) while neurotic-
ism (or emotionality) has its basis in the limbic system, reflected in lability
of the autonomic nervous system. Psychoticism appears to be related to
differences in androgen level and other hormonal secretions which are
themselves related to maleness.
However, as Claridge (1986) has well argued there are a number of
problems with this physiological account which have led other leading
workers in the field (e.g. Gray, 1982) to formulate a somewhat different
story. Gray's physiological account, which is finnly rooted in physiology,
although one based on rats rather than people, collapses neuroticism and
introversion into a continuum of anxiety which, to quote Claridge,
resembles a forty-five degree rotation of Eysenck's two factors, anxiety
running diagonally from neurotic introversion to stable extraversion.
However, Gray's system runs into difficulty with regard to P which he
cannot accurately place in factor spate.
The difficulty with the strict Eysenckian physiology lies in its complete
separation of central nervous system arousal and autonomic activation.

These, in reality, are unlikely to be entirely independent. This is what Gray
(1982) has sought to correct by incorporating both into a single dynamic
system.
Although there are problems with the physiological basis for these three
factors, there is some agreement that these physiological systems are in
some way involved. For example, despite the confusion of different
findings, arousal as indicated by the EEG is related to introversion-
extraversion (Gale and Eysenck, 1992). Some further support for this
general physiological approach comes from two sources. Since the mam-
malian nervous system is highly similar to that of human beings, there
should be some evidence of similar personality dimensions among mam-
mals. Eysenck (1976) argues that this is indeed the case. Thus Broadhurst
(1975) has studied emotionality in rats, a factor defined by faecal counts,
and Chamove et al. (1972) have claimed to have identified all three factors
in observations of rhesus monkeys.
If these factors are so intimately involved with such basic physiological
differences, then the heritability ratios for the factors should be high. As
has been fully discussed in Chapter 7 this is indeed the case, and even more
interestingly, between family environmental differences do not seem to
influence personality.
I shall not discuss the physiological basis of the Eysenckian factors
further. This is sufficient to indicate that Eysenck has attempted to root
these personality differences in biology, as is necessary for any adequate
theory of personality.
Relation ofE and N to conditioning and learning Eysenck (1967)

treats Nasa drive in the tradition ofHullian learning theory. Although he
admits that not all studies are in agreement and that there are some
anomalous results, he argues, as follows:
1 Emotion does act as a drive.
2 N refers to a dimension of greater emotional arousability.
3 High emotionality produces high drive in emotive situations.
4 Emotions acting as drives can hinder or facilitate performance depend-
ing on a number of interacting factors, including drive strength, task
difficulty, experience and other personality factors, all of which should
be included in any proper predictive equation.
There is one aspect of this formulation which does not seem to fit well
with the factor analytic evidence discussed earlier in this book. In the
studies of dynamic factor, anxiety or neuroticism did not appear as a drive.
It certainly affects learning but this may be as a temperamental rather than

a drive factor.
In this approach to learning theory, reinforcement is seen as drive
reduction, again in the Hullian tradition. However, drive reduction is
limited to reduction of N, which is seen as a reaction to threat. Thus
whatever reduces threat is reinforcing. This, as will be seen, is important
in the Eysenckian conception of neurosis and its treatment.
As regards extraversion, Eysenck integrates this factor into learning, thus
invoking excitation, which refers to cortical processes which facilitate
learning, conditioning, memory, perception and mental processing in
general, and the opposite of excitation inhibition, including satiation. This
is, essentially, a statement of the psychophysiological underpinning of
extraversion which was discussed in the previous section of this chapter.
From this Eysenck argued (1967) that introverts can be expected to
condition more quickly and that their responses would be more resistant
to extinction than extraverts, and that these differences would be greatest
when the experimental conditions were such that the build-up of inhibi-
tion was maximised. When this last condition was met, Eysenck (1967)
argued these predictions were supported although this qualification con-
veniendy eliminates results which do not fit the findings.
However, there is a severe objection to this theorising of a link between
conditioning and extraversion which was first made by Vernon (1963),
namely that for it to be confmned it would be necessary to show that
conditionability was a unitary factor, i.e. that the same variables affected in
the same way all conditioned responses. This is important because the
experimental evidence applies only to three responses - eyeblink condi-
tioning, GSR and verbal conditioning.
In the previous chapter it was seen that extraversion was related
significandy and negatively to academic success (clearly a form oflearning)
although correlations are generally modest. However, the results from
laboratory studies of the relation between extraversion and learning, as
Eysenck (1967) admits, are difficult to summarise, except to say that there
is some sort of significant relationship which might support the link
between poor learning and cortical inhibition. However, to be more
precise is difficult because there is so much variation depending upon
experimental conditions. Certainly it would seem dubious, in the extreme,
to extrapolate from these findings to real-life learning.
One feature of much of these results which does seem to be supported
in occupational psychology, for example, as has been argued by Smith
(1989), is that introverts are better at long and boring tasks, which tend to
produce inhibition, while extra~erts are better at brief and varied work.
Extraverts start better than introverts but, as inhibition builds up, begin to
deteriorate and make errors, although they may pick up again towards the
end. Extraverts are stimulus hungry and crave excitement, while introverts
prefer the opposite.
This brief discussion of how learning is woven into his psychometric
theory of personality indicates a number of severe difficulties with this
aspect ofEysenck's theorising. First it is clear that, although Eysenck can
explain them away, many results do not fit the theoretical predictions.
Secondly, as Claridge (1986) has argued, there are some internal difficulties.
Thus although arousal and activation are treated as separate properties (of
the CNS and the ANS), at times they are linked together, one ofthe reasons
that Gray (1982) attempted to modify physiological theories underpinning
these factors. In addition, as has been argued, until conditionability has
been shown to be unitary, the generalisability of the findings, even if they
are accepted, is dubious.
One of the strengths ofEysenck's theory (although not an intrinsic part
of it) is that it can be applied easily to real life behaviour, an extrapolation
which Eysenck has not been loath to execute, despite the cautions which
have been shown to be necessary.
As is well known, Eysenck has applied these theories to the analysis of
psychotherapy and has attempted to develop a tight scientific rationale for
behaviour therapy, based on classical conditioning theory. Classical con-
ditioning is seen as the basis for the development of neurotic symptoms,
as Eysenck (1982) argues. Cognitive concepts can be explained away in
these terms by suitable operational definitions. Thus introverts who con-
dition easily are likely to develop neurotic disturbances which are seen as
conditioned emotional responses to be extinguished in behaviour therapy.
However, as Lazarus (1986) has argued, such theorising is far too simplified
and simplistic for clinical use.
In his studies of smoking Eysenck (1980) has implicated his personality
factors in its determination and in the aetiology of disease, thus throwing
doubt on the simple linkage between lung cancer and cigarette smoking.
Similarly these factors have been implicated in sexual and marital satisfac-
tion (Eysenck, 1976) and in criminality (Eysenck, 1977; Eysenck and
Gudjonsson, 1989), although here there is the difficulty that criminality is
not a unitary concept. Eysenck and Gudjonsson (1989) try to argue from
the genetic influences which have been shown to be considerable for E,
N and P (see Chapter 7) and the known relationships of these factors with
criminality, that criminality might be expected to have similar hereditary
determinants. This case is indeed supported by the greater concordance for
criminal behaviour of MZ than DZ twins and by the fact that the
criminality of adopted children is detennined significandy by the crimi-

nality of their true parents. Indeed these authors show that estimates of
heritability for criminality are about the same as for personality - in the
region of 50-60 per cent. Again, this supports the influence of these
psychometric personality traits on highly important aspects of social
behaviour.
Conclusions It is clear from this discussion of Eysenck's theory of

personality that it is one which has incorporated, as central elements, the
factors which have emerged from his psychometric studies. His theory has
a biological basis and is applicable to a wide range ofbehaviour. Never-
theless, despite its apparent elegance and clarity, all is not as simple as
Eysenck would have us believe. There are too many experimental findings
which do not fit the hypotheses and neurosis is not a simple matter of
classical conditioning, just for example. However, as a basis for a theory
open to development as new data are collected, this theory is excellent and
it does what an adequate theory of personality must do - utilise the main
factor analytic findings.
The work of Cattell

In the final section of this chapter I shall discuss the theoretical account of
personality advocated by Cattell. This is undoubtedly the most complex,
quantified, psychometric theory of personality which has ever been de-
veloped and it is derived from more than fifty years' continuous
psychometry by Cattell and his colleagues.
The most comprehensive account of the theory is contained in the two
volumes of Cattell (1981), a book offormidable mathematical complexity.
A more simple description of the theory~ to be found in Cattell and Kline
(1977) and these are the best sources for an overall view of the theory,
although certain aspects are dealt with in more detail in other individual
publications, for example Cattell (1982) on the genetics of personality.
This summary account is derived from these sources in the main,
together with Cattell (1985), but is necessarily stripped of detail, for in fact
to describe the theory at all, huge numbers of researches and books have
to be taken into consideration.
As has been discussed throughout this chapter the aim of a good
psychometric theory of personality must be to make central the factors
which have been discovered th';ough factor analysis. For Eysenck this was
a relatively straightforward procedure since his work is concerned with
only three factors. For Cattell the position is entirely different. Thus, as has
been seen in previous chapters, Cattell has developed the following sets of
factors:
A Temperamental normal factors (measured by the 16PF and variants
designed for different age groups).
B Abnormal factors (measured in the CAQ, but perhaps less central to a
general theory of personality).
C Mood factors (measured by the Eight-State Questionnaire).
D Motivational Factors (measured by the CAT).
E Strength of interest factors (measured by the MAT).
All these factors and tests have been discussed in the relevant chapters of
this book on temperament and dynamics.
F Ability factors. Actually, in addition to these personality factors, Cattell
(e. g. 1971) has subjected the field ofabilities to factor analysis and ability
factors, which are beyond the purview of this book but are fully
discussed in its companion volume Intelligence: The Psychometric View
(Kline, 1991), form part ofhis theoretical structure.
Cattell is the doyen of psychometric personality theorists (and his approach
and arguments inform this book). Thus his theory has been developed on
the following principles:
1 Simple structure factor analyses are able, if the population of variables
is properly sampled, to identify the most important variables. Thus the
temperamental and dynamic factors which have been listed above must
form the basis of any theory of personality.
2 Such factors must have reliable and valid tests. These, in Cattell's theory,
are the tests listed above.
3 Psychometric personality theory has an underlying psychometric
model, namely that a particular action of any individual can be defined
by a specification equation which takes into account a subject's status
on the five sets offactors discussed above, and the ambient situation, all
of which are appropriately weighted for the particular act. If an action
can be predicted or specified by a specification equation then it can be
said to be understood. In this sense specification equations constitute
theories of particular actions and the better the fit, the better the theory.
As was discussed in Chapter 8, general specification equations for certain
jobs and for certain clinical syndromes have been developed by Cattell
and colleagues, using mainly the temperamental factors.
4 The development of all these factors needs to be studied so that factors
determining personality can be understood. It is not sufficient to state,
is possible that the linear algebra of this theory will have to be replaced by
some less elegant but more applicable mathematical procedures.
Conclusions In brief, therefore, it is argued that this Cattellian theory

constitutes a brilliant attempt to develop a proper psychometric theory of
personality. It is a superb blueprint for other researchers and his contribu-
tion to the statistical methods, especially factor analysis, which are integral
to such theorising, have been critically important and have been previously
discussed. What needs to be done now, is to improve the identification of
the factors and the psychometric statistical efficiency of the tests and to
continue Cattell's pioneering work in building up specification equations,
measuring situations and their interaction with factors, first efforts at which
are mentioned in Cattell (1982), and in developing the new algebra
necessary to accommodate the complexity of the equations. This is the task
which awaits the psychometrics of personality.
Chapter 10
In this chapter I shall summarise, as succincdy as possible, the main points

in the psychometric view of personality.
1 The psychometric view of personality is an example of a trait theory.
2 Underlying the psychometric view of personality is the psychometric
model ofhuman behaviour.
3 The psychometric model is a specification equation in which any
behaviour can be specified in terms of weighted scores on the main
personality and ability factors.
4 It is the task of psychometric personality research to detennine these
specification equations, both the variables and the weights.
5 To this end personality tests are required. Examination of the three types
- questionnaires or inventories, projective tests and objective tests -
showed that only inventories and objective tests were suitable for the
task.
6 Objective tests are not yet sufficiendy developed to be useful for more
than research and only factor analytic (because they are necessarily
unitary) personality inventories can yield psychologically meaningful
scores.
7 Factor analytic tests should be reliable and the factors need to be
externally validated.
8 Simple structure oblique factor analyses have been shown to yield
replicable and meaningful factors and the technical guidelines for
producing such factors - adequate sampling of variables and subjects,
the proper subject to variable ratio, choosing the correct number of
factors and a rotational procedure which maximises the hyperplane
count - are well established, as is the use of confirmatory analysis.
9 The factor analysis of temperamental personality tests has yielded five
factors, the 'big five' - extraversion, ·neuroticism or anxiety, tough-
mindedness, conscientiousness and open-mindedness.
Summary and conclusions 149
10 The first three of these are well established, especially in the work of
Eysenck and to a lesser extent Cattell, whose primary factors have
proved difficult to replicate. The other two factors are possibly better
subsumed as authoritarianism or obsessionality.
11 A study of other, temperamental personality inventories (not developed
by factor analysis) revealed no new factors. Criterion-keyed tests, such
as the MMPI, are useful only for screening.
12 Factoring in the abnormal sphere produced a number of factors ofwhich
the seven depression factors were of especial interest.
13 The study of dynamic factors has proved more difficult and there is less
agreement about results. Strictly to ensure factors are dynamic, R
analysis (of variables) is not satisfactory. Ideally P analysis (where the
same individual's scores on many occasions are factored) is required.
Short-term moods are difficult to capture.
As
14 regards moods or states, it is possible that there are only two: negative
and positive. However, subjects vary overall as to how moody they are.
15 Cattell has isolated seven strength of drive factors, of which three seem
critical: alpha (conscious id), beta (realised ego) and gamma (morality).
The significance of these factors is that in normal attitude scales, only
beta is tapped. Thus they are unlikely to be valid.
16 With the same test (MAT) Cattell measures ten ergs and sentiments,
ergs being basic biological drives and sentiments culturally-moulded
drives. However, there is little agreement concerning these factors in a
field which only Cattell has seriously investigated.
17 It is all these factors, together with ability factors which are beyond the
scope of this book, which must be the elements in the psychometric
view of personality and in the specification equations.
18 The heritability of three of the main personality factors, N, E and P,
has been investigated and there are substantial genetic contributions to
the population variance. Contrary to expectation between family
environmental variance is not influential. Similar findings have been
made with the big five.
19 There are substantial correlations between occupational membership
and success and the personality factors. Cattell indeed has attempted to
supply actual specification equations.
20 These factors have also proved useful in the prediction of educational
attainment and in clinical diagnosis. They could be valuable also in the
study of therapeutic treatment.
21 All these applied findings support the validity of the personality factors
and the psychometric view of personality.
22 None of the well-known theories of personality adequately deals with
150 Summary and conclusions
the personality variables shown to be salient by factor analysis. Thus

these factor analytic studies impugn their validity. Furthermore their
postulation of the importance of envirorunental facton in the develop-
ment of personality is probably wrong and certainly exaggerated.
Unique envirorunental experiences contribute to the envirorunental
vanance
23 Eysenck has developed a psychometric theory ofpenonality which links
his three £acton to arousal, inhibition and activation in the nervous
system and hence to conditioning. This theory has wide application to
neurosis, crime and marital adjustment, just for example.
24 Problems with the unitary nature of conditionability and the fit of the
theory to observations, especially in the sphere of clinical psychology,
have prevented its wide acceptance.
25 Cattell has developed an even more comprehensive and elaborate
theory of human behaviour which involves all his facton, including
ability facton. Utilising structured learning theory and the econetic
model he has attempted to draw up specification equations, even
including situations. However, the algebraic complexity and the prob-
lems over the identification of facton and the reliability of his tests,
renden the theory more of a brilliant blueprint than a substantive
contribution.
26 However, it is clear that by the better identification of facton in both
the temperamental and dynamic spheres and the elaboration of the
measurement of situations, a powerful psychometric view of penon-
ality, in the form of specification equations (not necessarily linear) for
a wide range ofbehavioun, can be drawn up.
References
Adler, A. (1927) Understanding Human Naturr. New York, Chilton.

Adorno, T.W., Frenkei-Bnmswick, E., Levinson, DJ. and Sandford, R.N. (1950)
1M Authoritarian Personality. New York, Harper & Row.
Allport, G.W. (1937) Personality: A Psychological Interpretation. New York, Holt,
Rinehart & Winston.
Altmeyer, B. (1981) Right-Wing Authoritarianism. Winnipeg, University of
Manitoba Press.
Amelang, M. and Borkenau, P. (1982) On the factor structure and external validity
of some questionnaire scales measuring dimensions of extraversion and
neuroticism (trans.) Zeitschriftfur Differentiate Diagnostiche Psychologie, 3, 119-146.
Arrindel, W.A. and van der Ende, J. (1985) An empirical test of the utility of the
observation-to-variables ratio in factor and components analysis. Applied
Psychological Measurement, 9, 165-178.
Ashkanasy, N.M. (1985) Rotter's internal-external scale. Confirmatory factor
analysis and correlations with social desirability for alternative scale formats.
Cited in Robinson, eta/. (1991).
Bandura, A. and Walters, R.H. (1963) Soda/ Learning and Personality Development.
New York, Holt, Rinehart & Winston.
Barrett, P. and Kline, P. (1981) The observation-to-variable ratio in factor analysis.
Personality and Group Behaviour, 1, 23-33.
Barrett, P. and Kline, P. (1982) Factor extraction: an examination of three methods.
Personality and Group Behaviour, 2, 94-98.
Barton, K. and Cattell, R.B. (1981) The Central State- Trait Kit (CTS): Experimental
Version. Champaign, Institute for Ability and Personality Testing.
Beck, A.T. (1962) Reliability of psychiatric diagnoses: a critique of systematic
studies. American]oumal of Psychiatry, 119, 21Q-215.
Bendig, A.W. (1959) Score reliability of dichotomous and trichotomous item
responses in the MPI.]oumal cifConsulting Psychology, 23, 181-185.
Block,). (1971) Lives Through Time. Berkeley, CA, Bancroft Books.
Blum, G.S. (1949) A study of the psychoanalytic theory of psychosexual
development. Genetic Psychology Monographs, 39, 3-99.
Bolton, B.F. (1986) Clinical diagnosis and psychotherapeutic monitoring. 348-376
in Cattell, R.B. and johnson, R.C. (eds) (1986).
Boyle, G J. (1989) Re-examination of the personality type factors in the Cattell,
152 References
Comrey and Eysenck scales: were the factor solutions by Noller tt al. optimal?
PmDMlity~~~tdlntlividual Difftm~m, 10, 1289-1299.
Briggs, K.C. and Myen, I.B. (1%2) 11tt Mym-Briggs Type Intlialtor. Princeton,
Educational Testing Service.
Broadhunt, P.L. (1975) The Maudsley reactive and non-reactive strains of rats- a
survey. Btluwioural Gtnttia, 5, 299-319.
Brown, R. (1965) Social Psyclwlogy. New York, Free Press.
Bryan, E. (1992) TwillS and Higher Multiple Births. A Guilt to their Naturr and Nurturr.
Sevenoaks, Edward Arnold.
Buck, J.N. (1970) 11tt House Trrt Pmon Technique: R.tvistd Manual. Los Angeles,
West Psychological Services.
Buros, O.K. (ed.) (1979) Eighth Mental Measurement Yeatbook. Highland Park,
Gryphon Press.
Butcher, J.N. (1990) MMPI-2 in Psyclwlogical Trratmmt. New York, Oxford
University Press.
Carroll, J.B. (1983) Individual differences in cognitive abilities. 213-235 in Irvine,
S.H. & Berry,J.W. (eds) (1983).
Cantain, G.S. (1957) 11tt Twice-&m: A Study ofa Community ofHigh Caste Hindus.
London, Hogarth Press.
Cattell, H . (1986) Clinical assessment by the 16PF, CAQ and MAT. 377-424 in
Cattell, R.B. &Johnson, R.C. (eds) (1986).
Cattell, R.B. (1957) Personality Motivation Structurr and Measurement. Yonken,
World Book Co.
Cattell, R .B. (1966) The Scree test for the number offacton. Multivariate Btluwiour
Research, 1, 14~161.
Cattell, R .B. (1971) Abilities: their Structure Growth and Action. New York,
Houghton Mifilin.
Cattell, R.B. (1973) Pmonality and Mood by Questionnairr. San Francisco,
Jossey-Bass.
Cattell, R.B. (1978) 11tt Scientific Use of Factor Analysis. New York, Plenum.
Cattell, R.B. (1981) Personality and !Laming 'llttory. New York, Springer.
Cattell, R.B. (1982) The lnhtritana of Personality and Ability. London, Academic
Press.
Cattell, R .B. (1985) Human Motivation and tht Dynamic Calculus. New York,
Praeger.
Cattell, R.B. and Bolton, L.S. (1969) What pathological dimensions lie beyond
the normal dimensions of the 16PF? A comparison ofMMPI and 16PF factor
domains.]oumal ofCoiiSulting and Clinical Psychology, 33, 18-29.
Cattell, R.B. and Butcher, H.J. (1968) The Prrdiction of Achitvtmtnt and Crrativity.
New York, Bobbs Merrill.
Cattell, R.B. and Child, D. (1975) Motivation and Dynamic Structurr. London, Holt,
Rinehart & Winston.
Cattell, R.B., Eber, H.W. and Tatsuoka, M.M. (1970) 11tt 16 Factor Personality
Qutstionnairr. Champaign, Institute for Ability and Penonality Testing.
Cattell, R.B. and Gibbons, B.D. (1%8) Penonality structure of the combined
Guilford and Cattell Penonality Questionnaires. Journal of Pmonality and Social
Psyclwlogy, 9, 107-120.
Cattell, R.B., Hom, J.L. and Sweney, A.B. (1970) Motivation Analysis Test.
Champaign, Institute for Ability and Penonality Testing.
References 153
Cattell, R.B. and johnson, R.C. (eds) (1986) FunctioruJ Psychological Testing. New
York, Brunner Maze!.
Cattell, R.B. and Kline, P. (1977) The Scientific Analysis tf PmoruJity and Motivation.
London, Academic Press.
Cattell, R.B. and Schuerger, J. (1976) The 0-A (ObjectiVt AruJytic) Test Battery.
Champaign, Institute for Ability and Penonality Testing.
Cattell, R.B. and Warburton, F.W. (1967) Objectivt PmoruJityand Motivation Tests.
Champaign, Univenity oflllinois Press.
Chamove, A.S., Eysenck, HJ. and Harlow, H.F. (1972) Penonality in monkeys:
factor analysis of Rhesus social behaviour. Quarterly journal tf Experimental
Psychology, 24, 496-504.
Child, D. (1991) Tht Essentials tf Factor Analysis (second edition). London, Holt,
Rinehart & Winston.
Christie, R. (1991) Authoritarianism and related constructs. 501-571 in Robinson,
J.P. tt al. (eds) (1991).
Claridge, G. (1985) Origins tf Mental Illness. Oxford, Blackwell.
Claridge, G.S. (1986) Eysenck's contribution to the psychology of penonality.
73-85 in Modgil, S. and Modgil, C. (cds) (1986) .
Comrey, A.L. (1970) The Comrey Personality Scales. San Diego, Educational and
Industrial Testing Service.
Cook, M. (1988) Personnel Selection and Productivity. Chichester, Wiley.
Cooper, C. and Kline, P. (1982) The internal structure of the Motivational Analysis
Test. British journal tfEducational Psychology, 52, 228-233.
Cooper, C. and McConnille, C. (1989) The factorial equivalence of the
state-extravenion positive affect and the state-anxiety negative affect. Personality
and Individual Differences, 10, 919-920.
Cooper, C. and McConnille, C. (1990) Interpreting mood scores: clinical
implications ofindividual differences in mood variability. BritishJournal tfMedical
Psychology, 63, 215-225.
Corah, N.L., Feldman, MJ., Cohen,I.S., Green, W., Meadow, A. and Rugwall,
E.A. (1958) Social desirability as a variable in the Edwards Penonal Preference
schedule. journal of Consulting Psychology, 22, 7Q-72.
Costa, P.T., Busch, C.M., Zondeman, A.B., Williams, R.B. and McCrae, R.R.
(1986) Correlations ofMMPI factor scales with measures ofthe five factor model
of penonality. Journal tf Personality Assessment, 50, 64o-650.
Costa, P.T. and McCrae, R.R. (1988) From catalogue to classification: Murray's
needs and the five factor model. Journal tf Personality and Social Psychology, 55,
258-265.
Costa, P.T. and McCrae, R.R. (in press). The Neo Pcnonality Inventory
(NEO-PI). In Briggs, S.R. and Check, J. (cds) (in press). Pmonality Measures
(Vol. 1). Grecnwich,JAI Press. ·
Costa, P.T., McCrae, R.R. and Holland, J.L. (1984) Pcnonality and vocational
interests in adulthood. journal tf Applied Psychology, 69, 39Q-400.
Costa, P.T., Zondeman, A.B., Williams, R .B. and McCrae, R.R. (1985) Content
and comprehensiveness in the MMPI: an item factor analysis in a normal adult
sample. Journal tf Personality and Social Psychology, 48, 925-933.
Cronbach, LJ. (1951) Coefficient alpha and the internal structure of tests.
Psychomttriluz, 16, 297-334.
Cronbach, LJ. (1976) Essentials tf Psychological Testing. New York, Harper & Row.
154 References
Cronbach, LJ. (1984) Bssmtials ofPsychologicdl Testing (revised edition). New Y orlt,
Harper & Row.
Cronbach, LJ. and Meehl, P.E. (1955) Construct validity in psychological tests.
Psydtologial Bulletin, 52, 177-194.
Curran, J.B. and Cattell, R.B. (1974) 1M Eight-Sta~ Questiotutaiw. Champaign,
Institute for Ability and Personality Testing.
Dahlstrom, W.G. and Walsh, G.S. (1960) An MMPI Ht~~~dboole. London, Oxford
Univenity Press.
Darlington, C.P. (1970) Hmdity, 25,655-656.
Digman, J.N. (1990) Pcnonality structure: emergence of the five factor model.
Annual Review of Psychology, 41,417-440.
Eaves, L.W., Martin, N.G. andEvscnclt, HJ. (1989) Gmts Cultuwand Personality.
London, Academic Press.
Edwards, A.L. (1969) Edwards Personal Prtjtm~a Schtduk. New York, Psychological
Corporation.
Eser, J.R. (1980) Cognitive Social Psychology. London, McGraw-Hill.
Exner, J.E. (1986) 1M Rorschach: A Compwhrnsive Sys~ (second edition). New
York, Wiley.
Eyscnck, H .J. (1959) The Ronchach, in Buros, O.K. (ed.) (1959) Fifth Mmtal
Measuwmmt Yealboole, Highland Park, Gryphon Press.
Eyscnck, HJ. (1967) 1M Biological Basis of Personality. Springfield, C. C. Thomas.
Eyscnck, HJ. (1976) Stx and Personality. London, Open Books.
Eyscnck, H .J. (1977) Crime and Personality. London, Roudedge & Kegan Paul
Eyscnclt, HJ. (1980) 1M Causes and Efficts of Smoking. London, Temple Smith.
Eyscnck, HJ. (1982) Neobehaviouristic (S-R) theory. 205-276 in Wilson, G.T.
& Franks, C .M. (eds) (1982) Con~mporary &haviour 1Mrapy. New York,
Guilford Press.
Eyscnck, HJ. (1989) Preface. In Friedman, A.F.t1 al. (1989).
Eyscnck, HJ. (1990) Genetic and environmental contributions to individual
differences: the three major dimensions ofpersonality.Joumal of Personality, 58,
245-261.
Eyscnck, HJ. (ed.) (1961) Handbook of Abnormal Psychology. New York, Basic
Books.
Eyscnck, HJ. (ed.) (1970) Readings in Extravmicm-lntrovmion . London, Staples
Press.
Eyscnck, HJ. and Eysenck, S.B.G. (1975) 1M Eysmcle Personality Questionnaiw.
Sevenoaks, Hodder & Stoughton.
Eyscnck, HJ. and Eysenck., S.B.G. (1969) Personality Structuw and Measuwmmt.
London, Roudedge & Kegan Paul.
Eyscnck, HJ. and Eyscnck., S.B.G. (1976) Psychoticism as a Dimension of Personality.
London, Hodder & Stoughton.
Eyscnck, HJ., Eysenck, S.B.G. and Barrett, P. (1992) 1M EPQR. Sevenoaks,
Hodder & Stoughton.
Eyscnclt, HJ, and Gudjonsson, G. (1989) Tht Causes and Cuws ofCriminality. New
York, Plenum Press.
Feldman, M. and Lewontin, R . (1975) The heritability hang up. Science, 190,
116~1168 .
Fenichcl, 0 . (1945) 1M Psychoanalytic 1Mory of Neurosis. New York, Norton.
References 155
Ferguson, G.A. (1949) On the theory of test development. Psychomttrilea, 14,

61-68.
Floderus-Myrhed, B., Pedersen, N . and Rasmuson, l. (1980) Assessment of
heritability for personality based on a short form of the Eysenck Personality
Inventory: A study of12,898 twin pairs. Bthallioural Gtnttics, 10, 153-162.
Fontana, D. (1981) Psychology for Ttachns. Leicester, British Psychological Society.
Freud, S. (1905) Three essays on sexuality. 135-243 in vol. 7 of'IM Standard Edition
of the Compkk Psychological works of Sigmund FrtUd. London, Tallistoclt Prtss and
the Institutt of Psyclwaru~lysis.
Freud, S. (1911) Psychoanalytic notes on an autobiographical account of a case of
paranoia (dementia paranoides). Vol. 3, 11, in Colkckd Psychological Works of
Sigmund Freud. London, Hogarth Press and Institute of Psychoanalysis.
Freud, S. (1933) New Introductory Lectures. Vol. 22 of 1M Standard Edition of the
Colltckd Psychological Works of Sigmund Freud. London, Hogarth Press and the
Institute of Psychoanalysis, 1966.
Freud, S. (1939) An Outline of Psychoaru~lysis. London, Hogarth Press and the
Institute of Psychoanalyis.
Friedman, A.F., Webb,J.T. and Lewak, R. (1989) Psychological Assessment with the
MMPI. Hillsdale, Erlbaum.
Fromm, E. (1965) The Heart ofMan: Its Gtnius for Good or Evil. London, Roudedge
& Kegan Paul.
Fulker, D.W. (1979) Nature and nurture: heredity. Chapter 5 in Eysenck, H.J. 1M
Structure and Measurement of Intelligence. New York, Springer-Verlag.
Gale, A. and Eyscnck, M. W . (cds) (1992) Handbook ofIndividual D!ffrrences: Biological
Perspectives. Chichester, Wiley.
Ghiselli, E.E. (1966) The Validity of Occupational Aptitude Ttsts. New York, Wiley.
Gillis,J.S. (1986) Classroom achievement and creativity. 447-465 in Cattell, R.B.
&Johnson, R.C. (eds) (1986).
Goldberg, L.R. (1983) The magical number five, plus or minus two: some
considerations on the dimensionality of personality descriptors. Research Paper,
Gerontology Research Center, Baltimore.
Gorsuch, R.L. (1974) Factor Aru~lysis. Philadelphia, Saunders.
Gorsuch, R.L. (1986) Measuring attitudes, interest, sentiments and values. 316-333
in Cattell, R.B. &Johnson, R.C. (eds) (1986).
Graham, J.R. (1990) MMPI-2 Assessing Pmonality and Pathology. New York,
Oxford University Press.
Gray, J A. . (1982) The Neuropsychology of Anxitty. Oxford, Clarendon Press.
Gruenbaum, A. (1984) The Foundations of Psychoaru~lysis. A Philosophical Critique.
Berkeley, University of California Press.
Grygier, T.G. (1975) Tht Dyru~mic Pmonality Inventory. Windsor, NFER.
Grygier, T.G. and Grygier, P. (1976) Manual to the Dynamic Pmonality Inventory.
Windsor, NFER.
Guilford, J.P. (1958) Psychometric Methods. New York, McGraw-Hill.
Guilford, J.P. (1959) Personality. New York, McGraw-Hill.
Guilford, J.P. (1967) Human Intelligence. New York, McGraw-Hill.
Guilford,J.S., Zimmerman, W.S. & Guilford, J.P. (1976) The Guilford-Zimmerman
Ttmptramrnt Survey Handbook. San Diego, EDITS.
Guthrie, G.M., Jackson, D.M., Astilla, R. and Elwood, B. (1981) Personality
measurement: do scales have different meanings in another culture? Kingston
156 References
(Ontario), paper at Nato Conference on Human Assessment and Cultural

factors.
Hakstian, A.R. (1971) A comparative evaluation of several prominent methods of
oblique factor transformations. Psychomttrilt4, 36, 175-193.
Hall, C.S. and Lindzey, G. (1957) Theories ofPmonality. New York, Wiley.
Hampson, S. and Kline, P. (1977) Personality dimensions differentiating certain
groups of abnormal offenders from non-offenders. British Journal of Criminology,
17, 31Q-331.
Harman, H.H. (1976) Modan Factor Analysis. Chicago, Univenity ofChicago Press.
Hathaway, S.R. and McKinley, J.C. (1951) Tht Minnmta Multiphasic Personality
Invtntory Manual (R.nrised). New York, Psychological Corporation.
Heim, A.W. (1975) PsychologiCAl Testing. London, Oxford University Press.
Helmes, E. (1989) Evaluating the internal structure of the Eysenck Personality
Questionnaire- objective criteria. Multivariak Behavioural &search, 24, 353-364.
Herriot, P. (ed.) (1989) Assessment and Selection in Organisations. Chichester, Wiley.
Holland, J.P. (1985a) The Holland Vocational Prefrrtnet Inventory. Odessa,
Psychological Assessment Resources.
Holland, J.P. ( 1985b) Making Carter Choim: A Thtory ofPersonality Types and Work
Environments. Englewood Cliffi, Prectice-Hall.
Holley, J.W. (1973) Rorschach analysis. 119-155 in Kline, P. (ed.) (1 973).
Holtzman, W.H. (1981) Holtzman Inkblot Technique. 47-83 in Rabin, A. I. (ed.)
(1981).
Hom,). and Knapp, JR. (1973) On the subjective character of the empirical base
of Guilford's structure of intellect model. Psychological Bulletin, 80, 33-43.
Howarth, E. (1976) Were Cattell's personality sphere factors correctly identified
in the first instance? British Journal of Psychology, 67, 213-230.
Howarth, E. (1980) TechniCAl Background and User Information for Stak and Trait
Inventories. Alberta, University of Alberta Press.
Hundleby, J.D. (1 973) The measurement of personality by objective tests. 65-87
in Kline, P. (ed.) (1973).
Irvine, S.H. and Berry,J.W. (eds) (1983) Human Assessment and Cultural Factors.
New York, Plenum.
Jackson, D.N. (1974) The Personality Research Form. New York, Research
Psychologists Press.
Jennrich, C.I. and Sampson, P.F. (1966) Rotation for simple loadings.
Psychometrika, 31, 313-323.
Jensen, A. (1980) Bias in Mental Testing. New York, Free Press.
Jewell, L.N. and Siegall, M. (1990) Contemporary Industrial/Organisational Psychology.
New York, West Publishing Co.
Jinks, J.L. and Fulker, J.W. (1970) Comparison of the biometrical, genetical,
MAVA and classical approaches to the study of human behaviour. PsychologiCAl
Bulletin, 73, 311-349.
John, O .P. (1990) The big five factor taxonomy: dimensions of personality in
natural language and in questionnaires. 66-100 in Pervin, L. (ed.) Handbook of
Personality Thtory ami Research. New York, Guilford Press.
Johnson, JR., Null, C ., Butcher,J.N. and Johnson, K.N. (1984) Replicated item
level factor analyses of the full MMPI. Journal of Personality and Social Psychology,
47, 105-114.
Joreskog, K.G. (1973) General methods for estimating a linear structural equation
References 157
system. In Goldberger, A.S. & Duncan, A.D. (eds) (1973) Structural EqU4tion
Mod~ls in tM Social Scimc~s. New York, Seminar Press.
Joreskog, KG. and Sorbom, D. (1979) Advanas in Factor Analysis and Structural
EqU4tion Models. Cambridge, MA, Abt Books.
Jung, C .G. (1910) The association method. Amtrican Journal of Psychology, 21,
219-269.
Jung, C.G. (1940) Th~ Integration oftM Pmottality. London, Routledge & Kegan
Paul.
Jung, C .G. (1949) Psychological Typts. London, Routledge & Kegan Paul.
Kameoka, V.A. (1986) The structure of the Clinical Analysis Questionnaire and
depression symptomatology. Multivariall' Behavioural Rmarch, 21, 105-121.
Kline, P. (1966) Extraversion, neuroticism and academic performance among
Ghanaian univenity students. BritishJournal ofEducational Psychology, 36, 93-94.
Kline, P. (1968) Obsessional traits, obsessional symptoms and anal erotism. British
Journal of Medical Psychology, 49, 299-305.
Kline, P. (1971) Ai3Q. Windsor, National Foundation for Educational Research.
Kline, P. (1975) The Psychology of Vocational Guidance. London, Batsford.
Kline, P. (1979) Psychometrics and Psychology. London, Academic Press.
Kline, P. (1981) Fact and Fantasy in Freudian Theory (second edition) . London,
Methuen.
Kline, P (1986) Handbook of Test Construction. London, Routledge and Kegan Paul.
Kline, P. (1988) Psychology Exposed: Or the Emperor's New Clothes. London,
Routledge.
Kline. P. (1991) Intelligence: The Psychometric View. London, Routledge.
Kline, P. (1992a) The Handbook of Psychological Testing. London, Routledge.
Kline, P. (1992b) Psychometric T~sting in Personnel &lection and Appraisal. Kingston,
Kroner.
Kline, P and Barrett, P. (1983) The facton in penonality questionnaires among
normal subjects. Advances in Behaviour Research and Therapy, 5, 141-202.
Kline, P. and Cooper, C. (1982) The internal structure of the Motivation Analysis
Test. British Journal of Educational Psychology, 52, 228-233.
Kline, P. and Cooper, C. (1984a) A factorial analysis of the authoritarian
penonality. British Journal of Psychology, 75, 171-176.
Kline, P. and Cooper, C. (1984b) A construct validation of the Objective Analytic
Test Battery (OATB). Personality and Individual Diffrrences, 5, 328-337.
Kline, P. (ed.) (1973) New Approaches in Psychological Measurement . Chichester,
Wiley.
Kline, P. and Grindley,}. (1974) A 28-day case study with the MAT.Journal of
Multivariate Exptrimental Personality and Clinical Psychology, 1, 13-32.
Kline, P. and Lapham, S. (1990) The PPQ. Exeter, University ofExeter.
Kline, P. and Lapham, S. (1991) The validity of the PPQ: a study of its factor
structure and its relation to the EPQ. Personality and Individual Differences, 12,
631--635.
Kline, P. and Storey, R. (1977) A factor analytic study of the oral character. British
Journal of Social and Clinical Psychology, 166, 317-328.
Kline, P. and Storey, R . (1978) The Dynamic Penonality Inventory: what does it
measure? British Journal of Psychology, 69, 375-383.
Kohn, P.M . (1972) The authoritarian-rebellion scale: a balanced F scale with
left-wing revenals. Sociometry, 35, 171-189.
158 References
Kreider, S. and Kreider, K. (1990) TM Cogniti~~r Foundations of Prrsonality Traits.

New York, Plenum.
Krug, S.E. (1 980) Clinical Analysis Qu~tionnairt. Champaign, Institute for Ability
and Penonality Testing.
Kuder, G.F. (1970a) Kudn- Gtnn-al Intmst Survty. Chicago, Science Research
Associates.
Kuder, G.F. (1 970b) Kuder Occupationa/Intmsts Survty. Chicago, Science Research
Associates.
Lacan, J. (1966) Ecrits. Paris, Seuil.
Layton, C. (1985) The relationship between externality and E, N, P and L: an
experiment and review. Prrsonality and Individual Differmces, 6, 505-507.
Lazarus, A.A. (1 986) On sterile paradigms and the realities of clinical practice:
critical comments on Eysenck's contribution to behaviour therapy. 247-257 in
Modgil, S. & Modgil, C. (eds) (1986).
Lefcourt, H . M~ (1991) Locus of control. 413--499 in Robinson, J.P. tt al. (eds)
(1991).
Likert, R.A. (1932) A technique for the measurement of attitudes. Archives of
Psychology, 140.
Loehlin, J.(1992) Genes and Environment in Personality Devtlopmmt. Newbury Park,
CA, Sage.
Loehlin, J.C. and Nichols, R.C. (1976) Heredity, Environment and Personality: A
Study of 850 Sets of Twins. Austin, Univenity of Texas Press.
Magnusson, D. and Endler, N .S. (eds) (1977) Personality at the Crossroads: Current
Issues in Intn-actional Psychology. Hillsdale, Erlbaum.
Martin, N. and Jardine, R. (1986) Eysenck's contribution to behaviour genetics.
13--47 in Modgil, S. and Modgil, C. (eds) (1986).
McClelland, D.C. (1961) Achieving Society. Princeton, Van Nostrand.
McCormick, E.J., Jeanneret, P.R . and Mecham, R .C. (1972) A study of job
characteristics and job dimensions as based on the Position Analysis
Questionnaire (PAQ).Journal of Applied Psychology, 56, 347-368.
McCrae, R.R. and Costa, P.T . (1985) Comparison ofEPI and psychoticism scales
with measures of the five factor theory of penonality. Personality and Individual
Diffmnces, 57, 17-40.
McCrae, R.R. and Costa, P.T . (1987) Validation of the five factor model of
penonality across instruments and observen. Journal of Personality and Social
Psychology, 52, 81- 90.
McCrae, R.R. and Costa, P. T . 1989a)( Rotation to maximise the construct validity
of facton in the NEO Penonality Inventory. Multivariate Behavioural Research,
24, 107-124.
McCrae, R.R. and Costa, P.T. (1989b) Reinterpreting the Myen-Briggs Type
Indicator from the penpective of the five factor model ofpenonality.Journa/ of
Personality, 57, 17-40.
McCrae, R.R, and Costa, P.T. (1990) Personality in Adulthood. New York,
·Guilford.
McCrae, R .R . and John, O .P. (in press) An introduction to the five-factor model
ofpenonality and its applications. journal of Personality (in press).
McDougall, W . (1932) Enagies ofMm. London, Methuen.
Mischel, W. (1968) Personality and Assmment. New York, Wiley.
References 159
Mischel, W. (1977) The intenction of person and situation. In Magnusson, D. &

Endler, N. S. (eds) (1977).
Mischel, W . (1984) Introduction to Pmonality (fourth edition). Tokyo, CBS.
Modgil, S. and Modgil, C. (eds) (1986) Hans Eysmclr: Consmsus and Controvmy.
London, The Falmer Press.
Murny, H .A. (1938) Explorations in Personality. Oxford, Oxford University Press.
Murstein, B.l. (1 %3) TMory and &uarch in Projecti~ Ttchniques. New York, Wiley.
Nesselroade,J .R. and Baltes, P .B. (1975) Higher-order convergence oftwo distinct
personality systems: Cattell's HSPQ and jackson's PRF. Multivariate &haviour
&search, 10, 387-408.
Noller, P., Law, H . and Comrey, A.L. (1987) Cattell, Comrey and Eysenck
personality factors compared: more evidence for the five robust factors. Journal
of Pmonality and Social Psychology, 53, 775-782.
Norman, W.T. (1963) Towards an adequate taxonomy of personality attributes.
Journal of Abnormal and Social Psychology, 65, 574-583.
Nunnally,J.O. (1978) Psychometric TMory. New Yqrk, McGraw-Hill.
Parker, C. and Kline, P. (1992) The psychological meaning of the VPI scales.
Pmonality and Individual Difftrencts (in press).
Pervin, L. (ed.) (1990) Handbook of Personality TMory and &search. New York,
Guilford Press.
Phares, E.J. (1976) Locus of Control in Personality. Morristown, General Learning
Press.
Phillipson, H . (1955) Object Relations Technique. London, Tavistock Publications.
Plomin, R. (1986) Development, Genetics and Psychology. Hillsdale, Erlbaum.
Popper, K. (1959) 1M Logic of Scientific Disco~. New York, Basic Books.
Rabin, A.I. (ed.) (1981) Assessment with Projective Techniques. New York, Springer.
Ray, J.J. (1970) The development and validation of a balanced dogmatism scale.
Australian Journal of Psychology, 22, 253-260.
Robinson, J.P., Shaver, P.R. and Wrightsman, L.S. (eds) (1991) Measures of
Personality and Social Psychological Attitudes. New York, Academic Press.
Rokeach, M. (1960) The Open and Closed Mind. New York, Basic Books.
Rorschach, H . (1921) Psychodiagnostics. Berne, Hans Huber.
Rose, R.J., Koskenvuo, M ., Kaprio,J., Sarna, S. & Langinvainio, N . (1988) Shared
genes, shared experiences, and similarity of personality.Journal of Personality and
Social Psychology, 164-171.
Rotter, J.B. (1966) Generalised expectancies for internal vs. external control of
reinforcement. Psychological Monographs, 80, No. 609.
Rotter,J.B. (1975) Some problems and misconceptions related to the construct of
external versus internal control ofreinforcement.]ouma/ ofConsulting and Clinical
Psychology, 43, 56--{;7.
Royce, JR. (1963) Factors as theoretical constructs. Chapter 24 in jackson, D.N.
& Messick, S. (eds) (1967) Problems in Human Assessment. New York,
McGnw-Hill.
Semeonoff, B. (1977) Projective Tests. Chichester, Wiley.
Shields, J. (1962) Monozygotic Twins. Oxford, Oxford University Press.
Shweder, R .A. (1975) How relevant is an individual difference theory of
personality? Journal of Personality, 43, 455-485.
Skinner, B.F. (1953) Science and Human Behaviour. New York, Macmillan.
160 References
Smith, M. (1989) Selection in high-risk and stressful occupations. 557-576 in

Herriot, P. (ed.) (1989).
Stone, W.F. and Lederer, G. (eds) (1991) Strmgth and Weakness: Tht Authoritarian
Pmonality Today. New York, Springer-Verlag.
Stricker, LJ. and Ross, R . (1964) An assessment of some structural properties of
the Jungian Personality Typology. ]oumal of Abnormal and Social Psychology, 68,
62-71.
Strong, E.K. and Campbell, D .P. (1974) Stron~Campbt/1 Interest lnwntory (Revised
edition). Stanford, Stanford University Press.
Strong, E.K., Campbell, D.P., Berdie, R .E. and Clerk, K.E. (1971) Strong Vocational
Intmst Blank. Stanford, Stanford Univenity Press.
Sweney, A.B., Anton, M.T. and Cattell, R .B. (1986) Evaluating motivation
structure, conflict and adjustment. 288-315 in Cattell, R.B. &Johnson, R.C.
(eds) (1986).
Sweney, A.B. and Cattell, R .B. (1980) Tht Vocational Intmst Measurr. Champaign,
Institute for Ability and Penonality Testing.
Tellegen, A., Lykken, D.T., Bouchard, T.J., Wilcox, K., Segal, N. and Rich, S.
(1988) Penonality similarity in twins reared apart and twins reared together.
journal of Personality and Social Psychology, 54, 1031-1039.
Thuntone, L.L. (1947) Multiple Factor Analysis: A Development and Expansion of
Vectol"3 of the Mind. Chicago, Univenity of Chicago Press.
Tupes, E.C. and Christal, R.E. (1961) Recurrent Penonality facton based on Trait
Ratings. USAF ASD Technical Report, Lackland, U .S. Air Force.
Velicer, W.F. (1976) Determining the number of components from the matrix of
partial correlations. Psychometrika, 41, 321-327.
Vernon, P.E. (1950) Tht Mecuurrmrnt of Abilities. London, Univenity of London
Press.
Vernon, P.E. (1963) PmonalityAssmmrnt. London, Methuen.
Vernon, P.E. (1964) Personality Assessment. London, Methuen.
Vetter, H.J. & Smith, B.D. (eds) (1971) Personality Theory: A Source &ole. New
York, Appleton-Century-Crofts.
Watson, D. & Tellegen, A. (1985) Towards a consensual structure of mood.
Psychological Bulletin, 98, 219-235.
Name index
Adler, A. 136 Cattell, H. 120, 121

Adorno, T. W. 75, 76 Cattell, R. B. : (1957) 7, 28, 40, 54,
.Allport, G. W. 24 57, 88, 108; (1966) 44; (1971) 83;
.Altmeyer, B. 76 (1973) 10, 13, 45, 52, 56, 61, 83,
Amelang 51, 57, 64 85, 86, 118; (1978) 7, 32, 38, 39,
Arrindel, W. A. 42 40, 43-7, 55, 56, 132; (1981) 7,
Ashkanasy, N. M. 78 52,53, 130,138,140, 146; (1982)
102, 106, 144, 146-7; (1985) 88,
Baltes, P. B. 74 90, 91, 121, 126, 134, 144; and
Bandura, A. 1, 2 Bolton (1969) 69; and Butcher
Barrett, P. :and Kline (1981) 42; and (1968) 127; and Child (1975) 89,
Kline (1982) 43, 44; Kline and 90, 92; and Gibbons (1968) 51;
(1983)40,43,50,56,60,62,66, and Hom and Sweney (1970) 119,
73, 75 120; and johnson (1986) 55, 88;
Barton, K. 87 and Kline (1977) 28, 40, 52, 85,
Beck, A. T. 124 86, 127, 140, 144; and Schuerger
Bendig, A. W . 18 (1976) 29, 58, 119, 120; and
Block, J. 138 Warburton (1967) 28, 29, 114;
Blum, G. S. 25, 134 Barton and (1981) 87; Curran and
Bolton, B. F. 69, 118, 119, 120 (1974) 84, 87; Sweney and (1980)
Borkenau 51, 57, 64 92, 93; tl al. (1964) 92; tl al. (1970)
Boyle, G. J. 57 30, 52, 55, 90, 108-9, 111; Eight
Briggs, K. C. 71, 136 State Questionnaire 84; factor
Broadhurst, P. L. 141 analysis of drives 88-96; factors
Broadway 72 52-8, 69, 74, 83, 149; motivational
Brown, R . 75 system 88-92, 137; personality
Bryan, E. 100 theory 144-7, 150; 16PF Test stt
Buck,J. N. 25, 27 16PF Test; work on moods and
Buros, 0. K. 24 states 85-8; work used in clinical
Butcher, H. J. 127 psychology 118-22
Butcher,). N. 71 Chamove, A. S. 141
Child, D. 32, 89, 92
Campbell, D. P. 94 Christal, R. E. 64
Carroll, J. B. 42 Christie, R. 76
Cmtairs, G. S. 26 Claridge, G. 123, 140, 143
162 Name index
Comrey, A. L. 18, 62-3 14, 17,40,50,56,58;Eysenck

Cook, M. 108 and (1976) 58, 59, 122, 140
Cooper, C.: and Kline (1982) 93,
121; and McConnille (1989) 84, Feldman, M. 101
87; and McConnille (1990) 84; Fenichel, 0. 88
Kline and (1982) 30; Kline and Ferguson, G. A. 16
(1984a) 75, 76, 135; Kline and Floderus-Myrhed, B. 103
(1984b) 30, 58, 120 Fontana, D. 127
Corah, N. L. 20 Freud, S.: (1905) 134; (1911) 25;
Costa, P. T.: and McCrae (1988) 64, (1933) 90, 134; (1939) 1; drives
74, 137; et al. (1984) 95; et al. 91, 126; personality theory 2,
(1985) 69, 124; et al. (1986) 64, 69; 134-5; personality variables 73, 79
McCrae and (1985) 57, 64; Friedman, A. F. 70, 124
McCrae and (1989a) 64, 72; Fromm, E. 1
McCrae and (1989b) 65; McCrae Fulker, D. W. 99, 100, 101, 102
and (1990) 64; McCrae and (in
press) 65 Gale, A. 141
Cronbach,L.J. lQ-11, 15,34,86 Ghiselli, E. E. 111
Curran, J. B. 84, 87 Gibbons (Cattell and, 1968) 51
Gillis, J. S. 129, 130
Dahlstrom, W . G. 22, 67, 123 Goldberg, L. R. 64
Darlington, C. P. 100 Gorsuch, R. L. 45, 91
Digman,). N. 35, 64, 74 Graham,). R. 67-8,71, 117, 123, 124
Gray,J. A. 14Q-1, 143
Eaves, L. W. 61, 103, 105--{), 138 Grindley,). 93
Edwards, A. L. 19, 20 Gruenbaum, A. 6
Eichmann, A. 75 Grygier, P. 72
Eiser,J. R. 3, 139 Grygier, T . G . 18, 72
Ende, van der J. 42 Gudjonsson, G. 143
Exner,J. E. 26, 125 Guilford,). P. : (1958) 42; (1959) 20,
Eysenck, H. J. : (1959) 25, 26; (1961) 44, 49; (1967) 47; et al. (1976) 49;
61; (1967) 58, 59, 6D-1, 122, 130, factors 49-52
140, 141, 142; (1970) 60; (1976) Guthrie, G. M . 74
122, 141, 143; (1977) 143; (1980)
33, 143; (1982) 143; (1989) 22, 67; Hakstian, A. R. 45, 56
(1990) 103,104,105, 139;and Hall, C. S. 1
Eysenck (1969) 50, 56; and Hampson, S. 27
Eysenck (1975) 14, 17, 40, 50, 56, Harman, H. H . 32, 42
58; and Eysenck (1976) 58, 59, Hathaway, S. R. 18, 67, 117
122, 140; and Gudjonsson (1989) Heim, A. W. 17, 37, 39
143; tl al. (1992) 59; clinical Helmes, E. 60
work 122-3; factors 58--<>2, 14Q-1, Herriot, P. 55, 108
150; on extraversion 35; Hitler, A. 138
personality theory 14Q-4, 150; see Holland,J. P. 64, 94
also Eysenck Personality Holley,). W. 27, 125
Questionnaire Holtzman, W. H. 27
Eysenck,M. W. 141 Hom,J. 47, 119, 120
Eysenck, S. B. G. : Eysenck and Howarth, E. 56, 83
(1969) 50, 56; Eysenck and (1975) Hull, C. 141, 142
Name index 163
Hundleby,J. D. 29 Lazarus, A. A. 143

Lederer, G. 75
Jackson, D . N. 23, 73, 137 Lefcourt, H . M. 77, 78
Jardine, R. 103 Lewontin, R. 101
Jennrich, C . I. 45 Likert, R. A. 18
Jensen, A. 129 Lindzey, G. 1
Jewell, L. N . 112, 113 Loehlin, J. C . 101, 103, 105
Jinks,]. L. 100, 101, 102
John, 0. P. 62, 64 McClelland, D. C. 126, 136
Johnson, R . C. 55, 69, 88, 124 McConnille, C . 84, 87
Joreskog, K. G. 43, 121 McCormick, E. J. 113
Jung, C. G.: (1910) 28; (1940) 1; McCrae, R . R. : and Costa (1985)
(1949) 71, 136; extraversion 35, 57, 64; and Costa (1989a) 64, 72;
136; personality theory 2, 71-2, and Costa (1989b) 65; and Costa
136;variabl~71-2, 79 (1990) 64; and John (in press) 62,
64; Costa and (1988) 64, 74, 137;
Kameoka, V. A. 53 Costa and (in press) 65
Kline, P. : (1966) 127; (1968) 134; McDougall, W .: (1932) 4, 81, 88,
(1971) 134, 145; (1975) 129; 90, 126, 130, 136; influence 88,
(1979) 44, 127; (1981) 7, 73, 89, 136-7; propensities 81, 91, 136-7;
134; (1986) 10, 21 ; (1988) 5, 6, 79; trait theory 4
(1990) 145; (1991) 39; (1992) 7, McKinley,]. C . 18, 67, 117
10, 13, 16, 21, 25, 27, 32, 45, 50, Martin, N. 103
51, 52,57, 58,65, 73, 74, 77,103, Meehl, P. E. 15
111-12, 114, 129; (1992a) 135; Mischel, W. 1, 3, 137-8
(1992b) 108, 114; and Barrett Modgil, C . 61
(1983) 40, 43, 50, 56, 60, 62, 66,
Modgil, S. 61
73, 75; and Cooper (1982) 30; and
Murray, H . A. : (1938) 4, 23, 24, 26,
Cooper (1984a) 75, 76, 135; and
73, 81, 90, 126; number of drives
Cooper (1984b) 30, 58, 120; and
91, 126, 137; PRF 71 , 73, 79;
Grindley (1974) 93; and Lapham
TAT 24; theory of needs 71, 73,
(1990) 43; and Lapham (1991) 65;
90, 136-7
and Storey (1977) 134; and Storey
Murstein, B.I. 26
(1978) 73; Barrett and (1981) 42;
Myers, I. B. 71, 136
Barrett and (1982) 43, 44; Cattell
and (1977) 28, 40, 52, 85, 86, 127,
140, 144; Cooper and (1982) 93, Ncsselroadc,J. R . 74
121; Hampson and (1977) 27; Nichols, R . C . 101, 103
Parker and (1992) 95 Noller, P. 57, 64
Knapp,]. R . 47 Norman, W. T. 64
Kohn, P.M. 76 Nunnally, J. 0 . 7, 10, 13, 37, 41-2,
Kreider, K. 4 47
Kreider, S. 4
Krug,S.E. 53, 69, 119,124 Parker, C . 95
Kuder, G . F.94 Pervin, L. 132
Phares, E. J. 77
Lacan,J. 1 Phillipson, H. 25
Lapham, S. 43, 65 Plomin, R . 98
Layton, C. 78 Popper, K. 3, 6, 130
164 Name index
Ray,J.J. 76 Stone, W. F. 75
Robinson,]. P. 23 Storey, R . 73, 134
Rokeach.~. 33,76 Stricker, L. J. 72
Rorschach, H. 24, 125 Strong, E. K. 93-4
Rose, R.J. 101, 103-4 Sweney, A. B. 91, 92, 93, 119, 120
Ross, R. 72
Rotter, J. B. 77-8 Tellegen, A. 84, 103
Royce, J. R. 37 Thatcher, ~- 138
Russell, B. 84 Thurstone, L. L. 40
Tupes, E. C. 64
Sampson, P. F. 45
Schuerger, J. 29, 58, 119, 120
Semeonoff, B. 25 Velicer, W. F. 44
Shields, J. 104 Vernon, P. E.: (1950) 14, 16; (1963)
Shwedcr, R . A. 138 142; (1964) 11' 13, 25, 26
Sicgall, ~ - 112, 113 Vetter, H.J. 72
Skinner, B. F. 2, 81
Smith, B. D. 72 Walsh, G. S. 22, 67, 123
Smith,~ - 142 Walters, R. H. 1, 2
Socrates 84 Warburton, F. W. 28, 29, 114
Sorbom, D. 121 Watson, D. 84
Subject index
ability: factors 145; traits 7 55,99-102

academic success 34, 55, 128, 129 Blacky Pictures 25, 134
acquiescence 19-20, 29 bloated specifics 61, 98
admission of wrongdoings 29 book titles 29
adolescents' career choice 95
aggression 2, 126, 135 career choice 95
agreeableness 64 CAQ see Clinical Analysis
Ai3Q 134-5 Questionnaire
alpha, 'conscious id' 89, 135, 149 CAT 145
alpha, Cronbach's 1Q-1 1 Central Trait-State Kit 87
ambiguous stimuli 24 chain P technique 86
anal character 134-5 change 2; factors 86
animal behaviour 139, 141 cheating 20, 114
anxiety: Cattell on 40; EPQ measures child: development 2; rearing 97, 98
14; factor analysis 35; heritability childhood experiences 2
97; measurement 29; physiology children: educational failure 127-8;
140; test hypotheses 15; trait and educational selection 128-9;
state 7, 82, 86, 87-8 projective tests for 25
appraisal 115-16 Children's Personality Questionnaire
arousal, cortical 140, 141, 150 52
assessment, forms of 11 Clinical Analysis Questionnaire
attitude tests 18, 90, 105 (CAQ) 53, 54, 69, 119, 124-5, 145
attribution theory 3, 138-40 clinical psychology 116-26
Australia, twin studies 103, 105 collective unconscious 2
authoritarian attitudes 105 communality 37-8
authoritarian person·ality, the 75--6, components, principal 42
79-80, 135 computer: programs 41, 42; reports
authoritarianism 149 from tests 115
concurrent validity 13-1 4
behaviour: learned 2; therapy 143 conditioning 141-4, 150
beta realised integrated interest 89, confidentiality 114-15
135--6, 149 confirmatory factor analysis 36, 46-7
big five factors 35--6, 63-5, 118, conscience 89, 135
133-4, 148, 149 conscientiousness 64
biometric: analysis 98, 106; methods 'conscious id' 89, 135, 149
166 Subject index
conservative attitudes 1OS environmental factors 97-106

construct validity 15, 55 EPQR59
construction, test 21-3 ergs 91, 135, 137, 146, 149
correlation(s): definition of term 33; ethical issues 114-15
magnitude 34-5; meaning 33--4; exploratory factor analysis 36
reproducing 38 extraversion: academic success
CPI 103 negative correlation 34; big five
criminality 14 3--4 factor 64; concept 132-3; E factor
criminals 27, 59 58,60,61, 105-~. 122,143, 149;
criterion keyed tests: construction EPQ measures 14, 60, 122, 140;
22-3, 123; use 113--14, 117, 124, Eysenck's learning theory 142-3;
149 factor analysis 35; Jungian 136;
Cronbach's alpha 10--11 state factor 87; temperamental trait
7, 81; test correlations 14; trait or
defence mechanisms 2 state 86, 87-8
delta, Ferguson's 16 extremes, endorsing 20
depression factors 54 exvia 86, see also extraversion
depressives 27 Eysenck Personality Inventory 58
depth psychologies 2 Eysenck Personality Questionnaire
depth psychometry 120 (EPQ) 14, 17, 42, 56, 58-{i1, 66,
deviation, standard see standard 103, see also EPQR
deviation
diagnosis, clinical 117 F scale 75, 76
differential validity 14 face validity 13
Direct Oblimin 45, 46, 48, 51 factor(s): big five 35-{i, 63-5, 118,
discriminatory power 16 133--4, 148, 149; Cattell 52-8,
distortion, deliberate 20, 31, 114 149; Comrey 62-3; definition of
distraction 29 37; Eysenck 58-{i2, 103, 150;
dogmatism 33 Guilford 50--1; heritability 97-8;
dR technique of factor analysis 85-7 identification 37, 40, 131; loadings
drives: definition 88; factor analysis of 37; MMPI 69-70; pattern 46;
88-91; structure of90--2; theories principal 42; rotation 38-9, 43-5;
81, 126, 136-7 scientific status 132-4; structure
Dynamic Personality Inventory (DPI) 46; validation 131
18, 72-3, 79 factor analysis: confirmatory 36, 4(r-7;
dynamic traits: distinct from definition oftemlS 32-3;
temperamental traits 81-3; drives description 35-{i; exploratory 36;
7, 81; factor analysis of83--4; states in test construction 22-3; probleOlS
and traits 82-3 in 39-41; technical difficulties 41;
technically sound 46
econetic model146, 150 factored personality tests in clinical
education, higher 34, 55, see also psychology 118
academic success faking20, 114
educational: psychology 116, 12(r-30; fascism 75
selection 128-9; theory 129-30 feedback 116
EEG 141 female/ male scores set' sex differences
ego 2, 89, 120, 149 Ferguson's delta 16
eigen values 38 Fidgetometcr 28-9
Eight State Questionnaire 84, 87, 145 Finland, twin studies 103
Subject index 167
forced-choice items 18-19 lability 60, 122, 140

latent roots 38
G analysis 27, 28 learning 141-3, 146
gamma 'superego' strength of interest lie scale items 20, 59
factor 89, 135, 149 like/ dislike item 18
genetic factors 55, 61, 97-106, 149 Likert scales 18
Guilford-Zimmerman Temperament locus of control 77-9
Survey 49
male/female scores see sex differences
health psychology 41 MAT see Motivation Analysis Test
heritability 61, 97-106, 149 matching method 109-10, 118
High School Personality Maudsley Medical Questionnaire
Questionnaire (HSPQ) 52, 74, 130 (MMQ) 58
Holland Vocational Preference Maudsley Personality Inventory 58
Inventory 64 MAYA method 102-3, 106
Holtzman Inkblot Technique 27 maximum likelihood analysis 43, 47
House Tree Person test (HTP) 25, 27 Maxplane 45, 63
hunger 7, 81-2 MBTI see Myers-Briggs Type
hypotheses 6, 8 Indicator
measurement: idiographic 17, 24;
nomothetic 17, 24; standard error
id 2; 'conscious' 89, 135, 149
of11-12
idiographic tests 17, 24
Mental Measurement Yearbooks 24
1-E scale 78
Minnesota Multiphasic Personality
incremental validity 14
Inventory (MMPI) 18, 22, 64,
individuation 2
67-70,79,117,123-5
inkblots 24-7, 125 MMPI-2 67, 71, 79, 117, 123-4, 149
intelligence 7, 14, 129, 145 moods see states
interest: measurement of88-90; motivation: clarifYing 83; factor
strength of89-90, 135-6; tests 14, analysis of drives 88-92; see also ergs
88,95-6 Motivation Analysis Test (MAT) 30,
introversion 140 90, 92-3, 119, 12Q-1, 130, 145,
introverts 55, 60, 142-3 148
intuitive analysis 108, 112 multiple analysis of variance see
inventories, personality 116-17, see MAYA
also questionnaires Myers-Briggs Type Indicator (MBTI)
ipsative scores 19, 114, 116 64, 71-2, 74, 79, 136
item analysis 21-2
item types 17-19 needs, theory of73, 74, 82, 137
NEO Personality Inventory 65-6, 95
Jackson Personality Research Form neurotic groups 55
see Personality Research Form neuroticism 40, 59, 6Q-l, 140; big
job analysis 112-13 five factor 64; N factor 58-9, 60,
job success, measuring 111 61, 105-6, 122, 141-2, 143, 149
neurotics 119, 120, 122-3
Kuder General Interest Survey 94 nomothetic tests 17, 24
Kuder Occupational Interest Survey norms16
94
Kuder Tests 94, 95 0 technique of factor analysis 48
168 Subject index
Object Relations Technique 25 Q technique of factor analysis 48

Objective Analytic Battery (OAB) quantification 5, 8
29-30,58,119,120 questionnaires, personality: advantages
objective tests 28-30, 31, 117 21, 30; construction 21-3; items
oblique rotations 44-5 17-19; problems 19-21,24, 30-1;
observations, numbers of 41 standardisation 23; validation 23
obsessional personality 75, 76, 79
obsessionality 33, 82, 149 R technique of factor analysis 47,
occupational choice 94 83-7, 149
occupational psychology 14, 55, 88, Rajputs, Rorschach study 26
107-16, 142 rating scales 18
opennness to experience 64 refutability 6
operant learning theory 2 regression method 110-11, 118
oral character 134-5 reinforcement 2, 81, 142, 146
orthogonal rotations 44 reliability: Cattell factors 54; Eysenck
factors 59; Guilford factors 50;
P technique of factor analysis 47, internal consistency 10-11; MAT
85-7, 149 92; measurement 10; MMPI 68;
parents 3, see also heritability NEO scales 65; test retest 11
persona 2 research design 6
personality: Cattell's theory 144-7, Right Wing Authoritarian Scale
150; Eysenck's theory 14Q--4, 150; (RWA) 76
psychometric model4, 5-7; trait Rorschach test 24, 26-7, 29, 117, 125
theory, 4 rotation offactors 38-9, 43-5;
Personality Research Form (PRF) 23, methods 45, 55
64, 73-5, 79, 82, 137 Rotoplot 45, 53
personnel selection 55, 95, 107-12 Rotter I-E scale 78-9
phobias 2
physiology 60-1, 140-1 samples 5
placebo 6 sampling 41, 109, 110, 118, 131
Position Analysis Questionnaire schizophrenics 27,119-20,122-3
(PAQ) 113 scientific method 3, 5, 8
PPQ 43,65 score, true 9-10
predictive validity 14 scorer, judgement of12, 26-7
Pre-school Personality Quiz 52 Scree test 44, 46, 48, 56
problem solving 7 self-knowledge 2
Procrustes rotations 47, 65 sentence completion techniques 25
projection 25 sentiments 91-2, 135, 149
projective tests 12, 24-8, 31, 117 sex differences 59, 61, 105
Promax 45, 51 sexuality 2, 126, 135
propensities 136-7 situationalism 3, 137-8
psychoanalytic theories 1-2, 25, R1, 16PF Test 42, 52, 54, 56-7, 64,
97,104,119-20,134-6 108-9, 111, 119, 124, 128, 129,
psychometric model of personality 4, 145
5-8 Slow Line drawing test 29
psychosexual forces 2 social desirability 19, 20, 78
psychoticism 59, 122, 140; P factor social learning theory 2-3, 77
58, 60, 61, 105-6, 122, 143, 149 social psychology 41
psychotics 23, 119-20, 122-3 Spearman-Brown prophecy formula 13
Subject index 169
standard deviation 12 trait theory of psychology 3, 4

standard error of measurement 11-12 traits, personality 4
standardisation, test 23 trichotomous forms 18
states: Cattell's work, 85-8; distinct true score 9-10
from traits 7, 82-3; factor analysis true/false item 18
of83--4; number 149 twin studies 61, 100--1, 103--4, 143
statistical analyses 5-6
stereotyping 138, 139 unconscious, the 1, 2
Strong-Campbell Interest Inventory United States of America (USA), twin
93--4,95 studies 103
Strong Vocational Interest Blank 93, universe of items 9
95 utilised cluster analysis 57
structured learning theory 146, 150
subjects: number of 41; ratio to
validation, test 23
variables 42; sampling 41
superego 2, 119, 130 validity: Cattell factors 55-7;
concurrent 13-14; construct 15,
'superego' strength of interest factor 89
Sweden, twin studies 103, 105 55; ditferential14; Eysenck factors
60--1; face 13; Guilford factors 50·
incremental14; MAT 93; MBTI'
target matrices 47
task analysis 112-13 72; NEO scales 65; personnel
TAT see Thematic Apperception Test appraisal testing 115-16; predictive
14
Tellegen scales 84, 87
temperament, study of 49 variable(s): definition of term 32;
temperamental traits 7, 81-3 numbers of 41; ratio of subjects to
test(s): comparison of types 30--1; 41,42
construction 21-3; discriminatory variance (definition of term) 33
power 16; idiographic 17, 24; item Varimax 45
types 17-19; nomothetic 17 24· Velicer test 44, 46, 48
norms 16; objective 28-30; ' ' vocational guidance 129
projective 24-8; questionnaires Vocational Interests Measure (VIM)
17-24; reliability 10--13; theory 92,93--4
9-11; validity 13-15 Vocational Preference Inventory
Thematic Apperception Test (TAT) (VPI) 94-5
24,25
trait-change factors 86-7 yes/no item 17-18

Personality - The Psychometric View

Uploaded by

Copyright:

Available Formats

Personality - The Psychometric View

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Personality - The Psychometric View

Uploaded by

Copyright:

Available Formats

The psychometric view of personality is well established but little dealt

Intelligence: the Psychometric View

The psychometric view

Simultaneously published in the USA and Canada

Reprinted 1994, 1997, 2004 and 2008

Routledge is an imprint of th e Taylor & Francis Group, an Informa business

© 1993 Paul Kline

Typeset in Bembo by Michael Mepham, Frome, Somerset

All rights reserved. No part of this book may be reprinted or reproduced or

British Library Cataloguing in Publication Data

Library of Congress Cataloging-in-Publication Data

ISBN 978-0-4 I 5-08978-4 (pbk)

1 The meaning of penonality 1

The meaning of personality

Different theories of penonality

method. Freudians stress the importance of defence mechanisms, such as

These will be entirely different from psychoanalytically based research.

Conclusions from descriptions ofpersonality theories

demonstrate that the definitions and concepts of personality are dependent

Trait theory of penonality

The psychometric model of personality

The psychometric model and scientific method

Essence of the scientific method

with the development of psychological tests thus ensuring as sound

The characteristics of good psychological tests

Universe of items Any set of items in a test is assumed to be a random

univene of items. This is, therefore, a notional score. However, as shall be

Internal consistency reliability Any measure, not simply a psycho-

Test-retest reliability There is no need of abstruse psychometric theory

Standard error of measurement

different levels of probability what the true score of an individual might

Importance of the standard error of measurement The standard

Factors influencing test reliability 1 am stressing the importance of

Face validity This refers to the appearance of a test. If it looks valid it

Concurrent validity To demonstrate concurrent validity, a test is

administered at the same time. For satisfactory concurrent validity a

Predictive validity A test is said to possess predictive validity if it can

Incremental and differential validity These two types of validity are

Construct validity Because, as has been seen, it is difficult to establish

Conclusions concerning validity It is clear from this discussion that

Summary and conclusions

TYPES OF PERSONAUTY TEST

Personality questionnaire or inventories

The True/False item I regularly feel sick before exams.

The Like/Dislike item Baked beans.

This type of item consists or words or phrases to which subjects have to

Items with rating scales People should be more self-controlled.

Various trichotomous forms These are variants of the Yes/No items

Forced-choice items These involve competing phrases of which sub-

Ipsative scores With forced-choice items in which each choice receives

Problems with personality questionnaires

Acquiescence This is the tendency to agree with responses, regardless

Social desirability This has already been mentioned in connection with

The tendency to endorse extremes Some subjects tend to endorse

The advantages ofpersonality inventories

The constrnction ofpersonality questionnaires

Item analysis in test construction The basis of test construction is to

Factor analysis in test construction This is highly similar to item

Validating the test