Psychological Medicine, 1989, 19, 513-518

Printed in Great Britain

PRELIMINARY COMMUNICATION

Development of a new self-report questionnaire to measure

self esteem
PHILIP ROBSON1
From the Department of Psychiatry, University of Oxford

SYNOPSIS In constructing a new self-report questionnaire for measuring self esteem, the aim has
been to satisfy psychometric requirements without sacrificing the intuitive meaning of the concept.
Self esteem is seen as a composite, and examination of its components may have more analytic or
predictive value than ' global' scores. Preliminary reliability and validity data are encouraging.

investigators have argued that it may in some

INTRODUCTION circumstances have a causal or maintaining role.
Self esteem is an idea rather than an entity and Doubts have often been cast on these claims for
the term signifies different things to different methodological reasons, but if reliable con-
people. Measuring instruments inevitably reflect firmation were available there might be im-
the theoretical stance of the inventor and portant practical implications for treatment.
sometimes intercorrelate poorly, suggesting that Pre-treatment levels of self esteem may, in certain
they are identifying different constructs, or circumstances, be an important predictor of
different elements of the same construct (Bridle, outcome (Fairburn et al. 1987).
1984; Lloyd et al. 1979). Wells & Marwell (1976) Thirdly, therapists of many different persua-
and Wylie (1974, 1979) have provided wide sions feel that improvement in self esteem is an
ranging critiques of construct definition, re- important element in the process of treatment,
search design and measurement techniques, and and this is a hypothesis which requires empirical
it is clear that most, if not all, of the existing confirmation.
means of measurement are in some way un- The inherent vagueness of the concept has led
satisfactory. to attempts to reduce and refine it, but these may
In the face of such difficulties, what is the result in measures whose content lacks intuitive
justification for attempting to pursue the concept credibility, or are, in the words of Ingham et al.
more effectively? First, it seems logical to (1986)'... conceptually primitive (though psycho-
suppose that a person's concern about the metrically sophisticated)'. This is the main
consequences of his actions must depend to shortcoming of the scale most widely used in this
some extent on the value he attaches to himself country, namely that devised by Rosenberg
and his existence. His opinion of himself is also (1965). Alternatively, self esteem has been
likely to influence his expectancies and explana- abandoned in favour of different constructs such
tion of what happens to him. It should therefore as self efficacy (Bandura, 1977).
be possible to predict certain aspects of be- To be useful in clinical research, a scale for
haviour from an accurate estimate of self measuring self esteem must be easily compre-
esteem. hensible to patients and quick to complete, and
Secondly, a review of the literature (Robson, demonstrate satisfactory psychometric proper-
1988) demonstrates that low self esteem has ties without sacrificing intuitive breadth of
been associated with a large number of undesir- meaning. No existing scale fulfils these require-
able traits, symptoms or behaviour, and some ments.
In attempting to quantify abstract notions
'Address for correspondence: Dr Philip Robson, University of
Oxford, Department of Psychiatry, Warneford Hospital, Oxford
some assumptions are required which are initi-
0 X 3 7JX. ally untestable. One assumption adopted by
513
PSM 19
514 P. Robson

several theorists (e.g. James, 1890; McDougall, Coopersmith (1967, JV=8) or Rogers &
1928; Rosenberg, 1965; Fleming & Courtney, Dymond (1954, N=l). Using a card-sort
1984) is that self esteem is based on beliefs technique, and having been given working
stemming from self evaluation of character, definitions for each component, nine experienced
abilities and behaviour. These beliefs gradually psychiatrists and psychologists were asked to
develop as a result of experiences during allocate each item to a component on face
childhood (Coopersmith, 1967) and later (Beck, validity, and to discard those items that did not
1967). Deficits in some of these aspects may be appear to fit any component. Items that did not
balanced by excesses in others, and if this is the achieve unanimous or near unanimous categor-
case it may be that shifts in the subordinate ization were rejected. Further items were dis-
elements which together give rise to self esteem carded on the basis of instability on test-retest,
have more analytic or predictive usefulness than heavily skewed response in a normative popu-
'global' self esteem scores. A scale which lation, or poor correlation with the mean overall
permitted examination of these individual com- score.
ponents would, therefore, constitute a logical This process of attrition resulted in a pool of
advance. 30 items (see Appendix) categorized as follows :
significance (5); worthiness (5); appearance/
DEFINITION OF SELF ESTEEM social acceptability (5); resilience and deter-
mination (5); competence (4); control over
The studies of Rosenberg (1965) have demon- personal destiny (4); value of existence (2). Since
strated the contribution of a sense of personal the relative scores of items and components is of
worthiness, appearance and social competence primary interest, a Likert scale was chosen for
to self esteem. Coopersmith (1967) has empha- scoring. The numerical scale for each item
sized the need for a feeling of competence and ranges from 0-7, giving a maximum score for
power, and Beck (1967) has pointed out the 'global' self esteem of 210. Four anchor points
importance of interpretation of events in ar- range from 'completely disagree' to 'completely
riving at a self view. Abramson et al. (1978) have agree'.
shown how attributional style may effect self The phrasing and ordering of the items were
esteem. arranged to minimize the commoner response
Drawing this work together, self esteem can errors (Oppenheim, 1966). Thus, there is a
be defined as follows: balance of positive and negative items to counter
' The sense of contentment and self acceptance the tendency to acquiesce, categories are well
that results from a person's appraisal of his own mixed to reduce the halo effect and the logical
worth, significance, attractiveness, competence, error, and double barrelled statements are
and ability to satisfy his aspirations.' avoided. The instructions for the questionnaire
A measuring instrument must aim to quantify requested a response ' as you typically feel most
this sense, or the individual elements of self of the time' in an attempt to capture trait, rather
appraisal. than state, attributes.

DERIVATION AND ARRANGEMENT OF SUBJECTS (See Table 1 for mean ages and
ITEMS FOR THE QUESTIONNAIRE sex ratios)
Based on the theoretical and empirical infor- Control group
mation reviewed elsewhere (Robson, 1988), Seventy people who showed no evidence of
seven components of self esteem were defined. psychological disorder were approached indi-
These consist of the subjective sense of sig- vidually by the author and asked to complete
nificance; worthiness; appearance and social the questionnaire. This was not a random
acceptablity; competence; resilience and deter- sample, but an attempt was made to obtain a
mination; control over personal destiny; and wide range of ages and occupational back-
the value of existence. grounds. The first 50 of these subjects also
Based on these components, 50 trait items completed the 10 items of the Rosenberg Self
were either devised (N = 35) or modified from Esteem Scale.
 Measurement of self esteem by questionnaire 515

Table 1. Mean age and sex ratio of the three (P < 0-0001) and mean stability of individual
samples items was 0-63 + 0-16 (s.D.).

Mean age Range Females (b) Patient groups

(years) S.D. (years) (%) In the anxiety sample split half correlation was
Control group 31 90 20-60 49 0-96 and the reliability coefficient derived from
Anxiety group 35 10 6 18-61 76 intraclass correlation was 0*91.
Psychotherapy 33 9-9 20-62 52
referrals Eighteen of the psychotherapy referrals were
sent the questionnaire by post following initial
contact and completed it again when they
attended for assessment, a mean interval of
The mean self esteem of this sample as
40 ± 18 (S.D.) days. Mean global scores at these
measured by the Rosenberg scale (10+1-2
two times were not significantly different
(standard deviation (s.D.)) did not differ sig-
(100-5 + 25 and 101-7 + 28 (s.D.) respectively,
nificantly from that reported in 'normal' popu-
giving a test-retest correlation of 0-88 (P
lations elsewhere (e.g. 1-1 + 1-4 (s.D.) Ingham
< 00001).
et al. 1986).
Anxiety group MEASURES OF VALIDITY
Fifty-one out-patients who met DSM-III criteria Convergent validity
for generalized anxiety disorder completed the
Rosenberg's (1965) self report measure of self
questionnaire. The first 24 subjects also com-
esteem has been widely used, and it is to be
pleted the 10 Rosenberg items.
expected that a new measure of self esteem
Psychotherapy group should correlate highly with this instrument.
This proved to be the case, with a positive
A consecutive group of 47 patients referred to correlation of 0-804 (P < 00001) being demon-
the Psychotherapy Department for assessment strated in the first 50 control subjects.
completed the questionnaire on first contact,
A similar comparison was carried out in the
and again at the assessment interview.
first 24 patients in the anxiety group, and here
the correlation was 0-85 (P < 00001).
MEASURES OF RELIABILITY Clinical validity
(a) Control group Nine experienced clinicians (four psychiatrists,
'Global' self esteem scores showed a normal three clinical psychologists, two psycho-
distribution with a mean of 137 + 2-41 (standard therapists) were asked to estimate the self esteem
error (s.E.)), giving 9 5 % confidence limits of of patients well known to them on a 10-point
132-142. Global scores did not relate to age (r visual analogue scale, and then to request the
= - 0 0 3 , P = 0-8) or sex (females = 135-4 + 21-5 patient to complete the questionnaire. Patients
(s.D.), males = 137-8+19-1 (s.D.)). Split half with organic or functional psychosis, acute
correlation was 0-931 (P < 0-0001) giving a mental disorder, or current substance abuse
reliability score of 0-96 using the Spearman- were excluded.
Brown formula. Thirty-seven patients were studied in this way.
Cronbach's coefficient alpha was 0-89. Reli- The correlation between the clinicians' estimate
ability coefficient, derived from intraclass cor- and the questionnaire was 0-7 (P < 00001).
relation of items and the total number of items
in the test (Nunnally, 1967), gave an identical Comparison with subjective assessment
value. The square root of this figure, 0-94, gives Twenty patients from the anxiety group who
the theoretical correlation of the test with 'true' said that they had heard of self esteem and
scores (Nunnally, 1967). understood the term were asked to estimate
Test-Retest stability was measured in a group their own self esteem on an 8-point visual
of 21 students assessed at intervals of at least 4 analogue scale. They then completed the ques-
weeks. Correlation of global scores was 0-87 tionnaire and the Beck Depression Inventory
19-2
516 P. Robson

(BDI). The correlation between the self evalua- therapy referrals were compared with the control
tion and the questionnaire was 07 (P = 00006). group, the former had significantly lower scores,
The correlation between the self evaluation and mean 'global' scores being 99-8±24 (S.D.) and
the BDI was -0-449 (P = 005). 137±20-2 respectively (P < 00001, / test).
When the effect of the BDI was partialled out,
the association between self evaluation and the
DISCUSSION
questionnaire remained significant (r = 0-624,
P < 001). This suggests that the relationship is The new self esteem scale was found to be
not mediated through the level of depression. acceptable to patients, and generally took only a
few minutes to complete. The reliability data
Discriminant validity detailed above compare favourably with that
reported for other self esteem measures (e.g.
Hypothesis Wylie, 1974). Preliminary investigations of con-
Low self esteem is generally held to be highly vergent and discriminant validity are also en-
associated with depressive disorder (e.g. Beck, couraging, and further assessment in larger
1967; Ingham et al. 1986) so that one would populations now seems appropriate.
predict a significant negative correlation between A factor analysis is being carried out in order
level of depression and self esteem score. In the to determine how the individual items inter-
anxiety sample the correlation between the BDI relate. The components upon which the scale is
and the new scale was -0-69 (P < 00001). based overlap, so that it might be expected, for
example, that 'significance' and 'worthiness'
Hypothesis may load together. This does not necessarily
A strong association between anxiety and low imply redundance since retaining breadth of
self esteem has frequently been reported (e.g. content is a primary aim of the 30-item format,
Rosenberg, 1962; Bagley et al. 1979). A group of which is intended to provide a rapid and
anxious patients should, therefore, be dis- systematic means of assessing self concept.
tinguishable from control subjects on the basis Responses to individual items, or patterns or
of self esteem score. When the anxiety group response, may reveal irrational, self defeating
was compared with the control group the attitudes or beliefs which are amenable to
mean global scores were 108 + 24-8 (S.D.) and modification by methods such as cognitive
137 + 20-2 respectively, a highly significant differ- behaviour therapy (Beck et al. 1979). The main
ence ( P < 0-0001, /test). function of the factor analysis is to provide the
basis for the construction of a shorter form of
Hypothesis the measure for research use, with items repre-
Anxious patients who have been successfully senting each component selected in order to
treated should demonstrate a significant improve- maximize reliability and validity.
ment in self esteem. Thirteen patients from the Self esteem is clearly a composite rather than
anxiety group whose anxiety ratings had normal- a single entity (Robson, 1988), and the main
ized following psychological treatment were purpose in adopting a component approach in
studied. Mean global scores were 109 + 39 (S.D.) devising the items was to ensure that the
before treatment and 166-4+17-2 afterwards, a questionnaire taps a broad view of the concept
significant improvement (P < 00001, paired t in a balanced way without overstressing any
test). single aspect. A second major motive stems from
my belief that in conditions which cause similar
Hypothesis changes in 'global' self esteem, for example
Many psychotherapists believe that low self anxiety and depression, the contribution made
esteem is a particular characteristic of patients by individual components may be quite different.
referred for dynamic psychotherapy (Bloch, Thus, in the same way that the presence or
1988). One would, therefore, predict that a absence of certain components of depressive
consecutive group of psychotherapy referrals disorder, such as diurnal mood variation or
should score significantly less on the self esteem weight loss, aid classification and give pointers
scale than control subjects. When the psycho- for choice of therapy, so might analysis of the
 Measurement of self esteem by questionnaire 517

components of self esteem give clues to aetio- I have a pleasant personality.

logical or maintaining factors and facilitate I never feel down in the dumps for very long.
more focused psychological treatment. Changes I often feel humiliated.
in component scores over time may be useful in I can usually make my mind up and stick to it.
investigating treatment process. Everyone else seems much more confident and
contented than me.
It is clear that abnormal mood states such as
Even when I quite enjoy myself there doesn't seem
depression distort many cognitive processes much purpose to it all.
including memory (Clark & Teasdale, 1982), I often worry about what other people are thinking
attributions and expectancies (Abramson et al. about me.
1978), and evaluation of the world and oneself There's a lot of truth in the saying: " what will be,
(Beck, 1967). Responses to self report question- will be."
naires are bound to be biased in these circum- I look awful these days.
stances and seem unlikely to represent accurately If I really try I can overcome most of my problems.
the targeted concept. This would be especially It's pretty tough to be me.
true of an abstract, subjective value judgement I feel emotionally mature.
such as self esteem. When people criticise me I often feel helpless and
second rate
For this reason it would seem sensible at this When progress is difficult, I often find myself
stage to concentrate on investigating self esteem thinking it's just not worth the effort.
and its components in circumstances in which I can like myself even when others don't.
acute disorders are not central. Interest in this Those who know me well are fond of me.
department is currently focused on its role in the
process of psychotherapy, its effect on pattern of REFERENCES
substance abuse and the outcome of treatment, Abramson, Lyn Y., Seligman, M. E. P. & Teasdale, J. D. (1978).
and the way its components correlate with Learned helplessness in humans: critique and reformulation.
Journal of Abnormal Psychology 87, 49-74.
outcome measures in chronic generalized anxiety Bagley, C , Verma, G. K., Mallick, K. & Young, L. (1979).
disorder. Personality, Self Esteem & Prejudice. Saxon House: England.
Bandura, A. (1977). Self efficacy: towards a unifying theory of
behaviour change. Psychological Review 84, 191-215.
I should like to thank Professor Michael Gelder for Beck, A. T. (1967). Depression: Clinical, Experimental and Thera-
advice and encouragement, and Pavlos Anastasiades peutic Aspects. Harper & Row: New York.
and Dr John Marzillier for their helpful comments on Beck, A. T., Rush, A. J., Shaw, B. F. & Emery, G. (1979). Cognitive
Therapy of Depression Guilford Press: London.
the manuscript. Bloch, S. (1988). Personal Communication.
Bridle, M. J. (1984). Comparison of the Tennessee self concept scale
and the Rosenberg Self Esteem Scale. Occupational Therapy
APPENDIX Journal of Research 4, 51-52.
The 30 items Clark, D. & Teasdale, J. D. (1982). Diurnal variation in clinical
depression and accessibility of memories of positive and negative
I am not embarrassed to let people know my experiences. Journal of Abnormal Psychology 91, 87-95.
Coopersmith, S. (1967). The Antecedents of Self Esteem. W. H.
opinions. Freeman: San Francisco.
I seem to be very unlucky. Fairburn, C. G., Kirk, J., O'Connor, M., Anastasiades, P. & Cooper,
I'm easy to like. P. J., (1987). Prognostic factors in bulimia nervosa. British Journal
If a task is difficult that just makes me all the more of Clinical Psychology 26, 223-224.
Fleming, J. S. & Courtney, B. E. (1984). The dimensionality of self
determined. esteem: II. Hierarchical facet model for revised measurement
There are lots of things I'd change about myself if scales. Journal of Personality & Social Psychology 46, 404-421.
I could. Ingham, J. G., Kreitman, N. B., Miller, P. M c C , Sashidharan, S. P.
I can never seem to achieve anything worthwhile. & Surtees, P. G. (1986). Self esteem, vulnerability and psychiatric
disorder in the community. British Journal of Psychiatry 148,
I don't care what happens to me. 375-385.
I have control over my own life. James, W. (1890) The Principles of Psychology. Macmillan:
Most people find me reasonably attractive. London.
Lloyd, C , Chang, A. F. & Powell, B. J. (1979). Relationship of
I'm glad I'm who I am.
different measures of self esteem to locus of control. Psychological
Most people would take advantage over me if they Reports 45, 160-162.
could. McDougall, W. (1928) An Outline of Psychology, 4th Edition.
I am a reliable person. Methuen: London.
Nunnally, J. M. (1967) Psychometric Theory. McGraw-Hill: New
It would be boring if I talked about myself. York.
When I'm successful there's usually a lot of luck Oppenheim, A. N. (1966). Questionnaire Design & Attitude Measure-
involved. ment. Heinemann: London.
518 P. Robson

Rogers, C. R. & Dymond, R. F. (1954). Psychotherapy and Person- Wells, L. E. & Marwell, G. (1976). Self Esteem: Its Conceptualisation
ality Change University of Chicaco Press: Chicago. and Measurement. Sage Publications: Beverley Hills.
Robson, P. i. (1988). Self esteem: a psychiatrist's view. British Wylie, R. C. (1974). The Self Concept, (Volume 1: a review of
Journal of Psychiatry 153, 6-15. methodological considerations and measuring instruments). Uni-
Rosenberg, M. (1962). The association between self esteem and versity of Nebraska Press: Lincoln.
anxiety. Journal of Psychiatric Research 1, 135-152. Wylie, R. C. (1979). The Self Concept, (Volume 2: Theory and
Rosenberg, M. (1965). Society and the Adolescent Self Image. research on selected topics). University of Nebraska Press: Lincoln.
Princeton University Press: Princeton, NJ.