Personality - The Psychometric View
Personality - The Psychometric View
Personality - The Psychometric View
Copyrighted Material
Other books by Paul Kline available from Routledge:
Copyrighted Material
Personality
Paul Kline
~~ ~~o~:~~n~~~up
LONDON AND NEW YORK
Copyrighted Material
First published I 993
by Routledge
27 Church Road, Hove, East Sussex BN3 2FA
Copyrighted Material
Contents
Copyrighted Material
Copyrighted Material
Chapter 1
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.
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
Copyrighted Material
2 The meaning of personality
Copyrighted Material
The meaning of personality 3
Copyrighted Material
4 The meaning of personality
Copyrighted Material
The meaning of personality 5
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.
Copyrighted Material
6 The meaning of personality
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
Copyrighted Material
The meaning of personality 7
Copyrighted Material
8 The meaning of personality
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.
Copyrighted Material
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.
True score The true score consists of the score of a subject on the
Copyrighted Material
10 Measurement of personality
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.
Copyrighted Material
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.
Copyrighted Material
12 Measurement of personality
Copyrighted Material
Measurement of personality 13
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.
Copyrighted Material
14 Measurement of personality
Copyrighted Material
Measurement of personality 15
Copyrighted Material
16 Measurement of personality
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.
Copyrighted Material
Measurement of personality 17
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
Copyrighted Material
18 Measurement of personality
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.
Copyrighted Material
Measurement of personality 19
Copyrighted Material
20 Measurement of personality
acquiescent individuals are not confused with high scoren on the test.
Actually, in test construction items which attract acquiescent responses can
be eliminated.
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.
Copyrighted Material
Measurement of personality 21
Copyrighted Material
22 Measurement of personality
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.
Copyrighted Material
Measurement of personality 23
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.
Copyrighted Material
24 Measurement of personality
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,
Copyrighted Material
Measurement of personality 25
Copyrighted Material
26 Measurement of personality
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
Copyrighted Material
Measurement of personality 27
Copyrighted Material
28 Measurement of personality
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.
Copyrighted Material
Measurement of personality 29
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
Copyrighted Material
30 Measurement of personality
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.
Copyrighted Material
Measurement of personality 31
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.
Copyrighted Material
Chapter 3
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.
Copyrighted Material
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.
Copyrighted Material
34 Factor analysis
Copyrighted Material
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,
Copyrighted Material
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.
Copyrighted Material
Factor analysis 37
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.
Copyrighted Material
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.
Copyrighted Material
li,
Factor analysis 39
'F
I
I
----- a, =. 5 . •.. :
I
~ ······."!-· a, = .29
a2 = - .57 1
-a-2-=--~.4~~~0~--------F 2
Copyrighted Material
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
Copyrighted Material
Factor analysis 41
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.
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.
Copyrighted Material
42 Factor analysis
Copyrighted Material
Factor analysis 43
Copyrighted Material
44 Factor analysis
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.
Copyrighted Material
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.
Copyrighted Material
46 Factor analysis
Copyrighted Material
Factor analysis 47
P technique Here test scores obtained from one individual are factored.
Copyrighted Material
48 Factor analysis
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.
Copyrighted Material
Chapter 4
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.
Copyrighted Material
50 The factor analysis of temperament
Copyrighted Material
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.
Copyrighted Material
52 The factor analysis of temperament
Copyrighted Material
The factor analysis of temperament 53
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.
Copyrighted Material
54 The factor analysis of t_
emperament
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,
Copyrighted Material
The factor analysis of temperament 55
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.
Copyrighted Material
56 The factor analysis of temperament
Copyrighted Material
The factor analysis of temperament 57
Copyrighted Material
58 The factor analysis of temperament
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.
Copyrighted Material
The factor analysis of temperament 59
Copyrighted Material
60 The factor analysis of temperament
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.
Copyrighted Material
The factor analysis of temperament 61
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.
Copyrighted Material
62 The factor analysis of temperament
political allegiance and school learning (see Modgil and Modgil, 1986, for
interesting summary discussions of many of these and similar issues).
Copyrighted Material
The factor analysis of temperament 63
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.
Copyrighted Material
64 The factor analysis of temperament
Copyrighted Material
The factor analysis of temperament 65
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.
Copyrighted Material
66 The factor analysis of temperament
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.
Copyrighted Material
Chapter 5
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.
Copyrighted Material
68 Other types of psychometric tests
Copyrighted Material
Other types of psychometric tests 69
Copyrighted Material
70 Other types of psychometric tests
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.
Copyrighted Material
Other types of psychometric tests 71
Copyrighted Material
72 Other types of psychometric tests
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.
Copyrighted Material
Other types of psychometric tests 73
Copyrighted Material
74 Other types of psychometric tests
Copyrighted Material
Other types of psychometric tests 75
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.
Copyrighted Material
76 Other types of psychometric tests
Copyrighted Material
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
Copyrighted Material
78 Other types of psychometric tests
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.
Copyrighted Material
Other types of psychometric tests 79
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.
Copyrighted Material
80 Other types of psychometric tests
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.
Copyrighted Material
Chapter 6
Personality dynamics
The psychometric view
Copyrighted Material
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.
Copyrighted Material
Personality dynamics 83
Copyrighted Material
84 Personality dynamics
Copyrighted Material
Personality dynamics 85
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.
Copyrighted Material
86 Personality dynamics
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.
Copyrighted Material
Personality dynamics 87
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.
Copyrighted Material
88 Personality dynamics
of this chapter, also act as drives and this accounts for the emergence of
trait-change factors.
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
Copyrighted Material
Personality dynamics 89
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:
Copyrighted Material
90 Personality dynamics
Copyrighted Material
Personality dynamics 91
Copyrighted Material
92 Personality dynamics
Reliability and validity of the MAT There is little doubt that the
Copyrighted Material
Personality dynamics 93
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.
Copyrighted Material
94 Personality dynamics
Copyrighted Material
Personality dynamics 95
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.
Copyrighted Material
96 Personality dynamics
Copyrighted Material
Chapter 7
Heritability of personality
Copyrighted Material
98 Heritability of personality
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.
Copyrighted Material
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
Copyrighted Material
100 Heritability of personality
Copyrighted Material
Heritability of personality 101
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-
Copyrighted Material
102 Heritability of personality
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.
Copyrighted Material
Heritability of personality 103
Copyrighted Material
104 Heritability of personality
Copyrighted Material
Heritability of personality 105
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).
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
Copyrighted Material
106 Heritability of personality
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.
Copyrighted Material
Chapter 8
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.
Copyrighted Material
108 Personality testing in applied psychology
Copyrighted Material
Personality testing in applied psychology 109
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
Copyrighted Material
11 0 Personality testing in applied psychology
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.
Copyrighted Material
Personality testing in applied psychology 111
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.
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
Copyrighted Material
112 Personality testing in applied psychology
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.
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.
Copyrighted Material
Personality testing in applied psychology 113
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.
Copyrighted Material
114 Personality testing in applied psychology
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.
Copyrighted Material
Personality testing in applied psychology 115
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.
Copyrighted Material
116 Personality testing in applied psychology
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.
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.
Copyrighted Material
Personality testing in applied psychology 117
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.
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.
Copyrighted Material
118 Personality testing in applied psychology
Copyrighted Material
Personality testing in applied psychology 119
Copyrighted Material
120 Personality testing in applied psychology
Copyrighted Material
Personality testing in applied psychology 121
Copyrighted Material
122 Personality testing in applied psychology
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.
Copyrighted Material
Personality testing in applied psychology 123
Copyrighted Material
124 Personality testing in applied psychology
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
Copyrighted Material
Personality testing in applied psychology 125
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.
Copyrighted Material
126 Personality testing in applied psychology
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.
Copyrighted Material
Personality testing in applied psychology 127
Copyrighted Material
128 Personality testing in applied psychology
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.
Copyrighted Material
Personality testing in applied psychology 129
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
Copyrighted Material
130 Personality testing in applied psychology
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.
Copyrighted Material
Chapter 9
Copyrighted Material
132 Personality theory
Copyrighted Material
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.
Copyrighted Material
134 Personality theory
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.
Copyrighted Material
Personality theory 135
Copyrighted Material
136 Personality theory
Copyrighted Material
Personality theory 137
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
Copyrighted Material
138 Personality theory
Attribution theory
In our discussion of situationalism one of the aq,>Uments utilised by
Copyrighted Material
Personality theory 139
Copyrighted Material
140 Personality theory
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.
Copyrighted Material
Personality theory 141
Copyrighted Material
142 Personality theory
Copyrighted Material
Personality theory 143
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
Copyrighted Material
144 Personality theory
Copyrighted Material
Personality theory 145
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,
Copyrighted Material
Personality theory 147
is possible that the linear algebra of this theory will have to be replaced by
some less elegant but more applicable mathematical procedures.
Copyrighted Material
Chapter 10
Copyrighted Material
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
Copyrighted Material
150 Summary and conclusions
Copyrighted Material
References
Copyrighted Material
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.
Copyrighted Material
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.
Copyrighted Material
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.
Copyrighted Material
References 155
Copyrighted Material
156 References
Copyrighted Material
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.
Copyrighted Material
158 References
Copyrighted Material
References 159
Copyrighted Material
160 References
Copyrighted Material
Name index
Copyrighted Material
162 Name index
Copyrighted Material
Name index 163
Copyrighted Material
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
Copyrighted Material
Subject index
Copyrighted Material
166 Subject index
Copyrighted Material
Subject index 167
Copyrighted Material
168 Subject index
Copyrighted Material
Subject index 169
Copyrighted Material
Copyrighted Material