Psychological Medicine, 1982, 12, 871-878

Printed in Great Britain

The Eating Attitudes Test: psychometric features and

clinical correlates
DAVID M. GARNER, 1 MARION P. OLMSTED, YVONNE BOHR
AND PAUL E. GARFINKEL
From the Clarke Institute of Psychiatry and the University of Toronto, Canada

SYNOPSIS Psychometric and clinical correlates of the Eating Attitudes Test (EAT) are described
for a large sample of female anorexia nervosa (N = 160) and female comparison (TV = 140) subjects.
An abbreviated 26-item version of the EAT (EAT-26) is proposed, based on a factor analysis of
the original scale (EAT-40). The EAT-26 is highly correlated with the EAT-40 (r = 0-98) and the
three factors form subscales which are meaningfully related to bulimia, weight, body-image variables
and psychological symptoms. Whereas there are no differences between bulimic and restricter
anorexia nervosa patients on the total EAT-26 and EAT-40 scores, these groups do indicate"
significant differences on EAT-26 factors. Norms for the anorexia nervosa and female comparison
subjects are presented for the EAT-26, EAT-40 and the EAT-26 factors. It is concluded that the
EAT-26 is a reliable, valid and economical instrument which may be useful as an objective measure
of the symptoms of anorexia nervosa.

INTRODUCTION the HSCL (Derogatis et al. 1974), in samples of

anorexic patients and dance students. Button &
Increasing interest in the objective measurement
Whitehouse (1981) administered the EAT to a
of the symptoms and psychophysiological fea-
large sample of students at a College of
tures of anorexia nervosa is reflected in the
Technology and reported that the instrument
development of several rating scales over the past
was useful in identifying a group with abnormal
decade. Slade (1973) originally proposed an
concerns with eating and weight. Thompson &
observer rating scale for assessing anorexic
Schwartz (1982) used the EAT to identify college
behaviour patterns in an in-patient setting.
women with abnormal eating concerns who were
Goldberg et al. (1980) developed a 63-item
later distinguished from anorexic patients in
self-report measure of anorexic attitudes and
terms of social adjustment. Hawkins & Clement
behaviour, also designed strictly for use in
(1980a) organized items from the EAT into
hospital. Fichter & Keeser (1980) have devised
subscales measuring dieting and food preoccupa-
a self-rating scale for assessing several aspects of
tion; however, subscales were not empirically
anorexic behaviour; however, validation was not
derived or validated. We have found that anorexic
reported for English speaking subjects.
subjects who markedly overestimate their body
The Eating Attitudes Test (EAT) has been size on an operational measure of body-image
proposed as an objective, self-report measure of (Garner et al. 1976) have significantly higher
the symptoms of anorexia nervosa (Garner & EAT scores than those who moderately over-
Garfinkel, 1979). It has been used as a screening estimate or underestimate their size (Garner &
instrument for detecting previously undiagnosed Garfinkel, 1981). Hood et al. (1982) reported
cases of anorexia nervosa in populations at high that high EAT scores were related to external
risk for the disorder (Garner & Garfinkel, 1980). locus of control for a group of anorexia nervosa
Garner & Garfinkel (1980) found that EAT patients.
scores were independent of weight but were
related to emotional disturbance, as measured by Although the initial reliability and validity of
1
the EAT have been reported (Garner &
Address for correspondence: Dr David M. Garner. Clarke Garfinkel, 1979), the relationship between symp-
Institute of Psychiatry. 250 College Street. Toronto. Ontario M5T
1R8. Canada. tom areas and clinical features of anorexia
871
872 D. M. Garner, M. P. Olmsted, Y. Bohr and P. E. Garfinkel

nervosa has not been established. The total EAT FC group was 20-3 (S.D. = 2-7) and they had a
score is derived from a heterogeneous item pool, mean of 96-6% of average weight for age and
and while it may indicate the overall level of height (Health and Welfare Canada, 1954).
symptoms it does not provide item clusters
which may relate to clinical characteristics of Procedures
interest. Moreover, while the entire scale or Following an initial consultation, the EAT and
various items have been employed in recent the instruments listed below were administered
studies, questions regarding psychometric quali- to the AN patients. The number of AN subjects
ties, including factor structure and short-forms, receiving individual tests varied, since the
have been raised. The purpose of this paper is to psychometric battery administered to the sub-
describe a factor analysis of the EAT on a large jects has changed over time. However, for each
sample of anorexic patients and to determine measure below there were no significant EAT
whether item clusters are associated with clinical score differences between the group for whom
and personality features. data were available and the group for whom data
were unavailable. The FC subjects only completed
the EAT. The EAT was scored according to the
METHOD procedure described by Garner & Garfinkel
(1979).
Subjects The Distorting Photograph Technique (DPT)
The anorexia nervosa (AN) sample consisted of has been employed as one measure of body-image
160 female patients who were consecutive (Garner et al. 1976; Garfinkel et al. 1979;
consultations seen at the Clarke Institute of Strober, 1981). This allows the subject to
Psychiatry. Diagnosis was made according to a estimate her body width using a photograph
modified version of the Feighner et al. (1972) which can be distorted along the horizontal axis.
criteria (Garfinkel & Garner, 1982). The mean Both self-estimates and ideal estimates were
age of the sample was 21-5 years (s.D. = 5-4). The obtained.
sample was heterogeneous in that patients were Body Dissatisfaction Scale (BDS) is an
at various stages of illness when tested, although adapted version of the Berscheid et al. (1973)
none could be considered recovered. Approxi- scale. It consists of a list of 18 body parts, and
mately one half (48%) of the patients were of the the subject is instructed to indicate her degree of
'restricter' subtype and the remainder had the satisfaction with her own body on a 6-point scale
complication of bulimia. Bulimia was defined, (extremely, quite or somewhat satisfied, some-
following Garfinkel et al. (1980, p. 1037), as what, quite or extremely dissatisfied). Dissatis-
' episodes of excessive ingestion of large quantities faction on this scale has been found to correlate
of food that the patient viewed as ego-alien and with overestimation on the DPT (r = 0-48,
beyond her control'. The mean duration of P < 0007) and a measure of self-esteem
illness for the entire sample was 4-3 years, and (r=-0-65, P < 0-001) in a small sample of
there were no differences between the bulimic anorexic patients (Garner & Garfinkel, 1981).
and restricter subtypes in age or duration of A body-image composite score. Body-image is
illness. While the entire sample was 741 % of a multidimensional concept which has been
average weight for age and height (Health and operationalized in different ways. In anorexia
Welfare Canada, 1954), the bulimics were nervosa, body-image disturbances have been
heavier than the restricters (80-2% v. 67-5% of defined as an overestimation of one's own size,
average weight: t = 5-24, P < 0001). preference for a 'thin' ideal image, and negative
The female comparison (FC) group consisted attitudes towards one's body (Garner & Gar-
of 140 female university students from first- and finkel, 1981). All of these aspects of body-image
second-year psychology courses. They were were incorporated into a body-image composite
asked to complete the questionnaire during a score which was calculated for each subject by
regular class session; participation was voluntary summing the standardized DPT self-estimate,
and 98% of the subjects approached agreed to DPT ideal-estimate (after reversing the sign so
complete the questionnaire. The mean age of the that a thin ideal reflects disturbance) and BDS
 The Eating Attitudes Test 873

scores. Thus a high score indicated body-image 14, 25) loaded approximately equally on Factors
disturbance from an aggregate of measures (i.e. I and II and, after inspection of the item content,
self-overestimation, thin ideal, and high BDS). were retained only on Factor I.
Locus of Control Scale is a 3 factor scale, Items loading on the first factor, labelled
described by Reid & Ware (1973). The rationale 'dieting', relate to an avoidance of fattening
for its use in studies of anorexia nervosa was foods and a preoccupation with being thinner.
described previously by Garner et al. (1976). The second factor,' bulimia and food preoccupa-
Hopkins Symptom Check List (HSCL) is a tion', consists of items reflecting thoughts
58-item self-report index yielding subscores for about food as well as those indicating bulimia.
depression, anxiety, interpersonal sensitivity, Items on the third factor, 'oral control', relate
obsessionality, and somatization (Derogatis et to self-control of eating and the perceived
al. 1974). Results with the HSCL have previously pressure from others to gain weight. Factors I
been reported for patients with anorexia nervosa and II are polar opposites and seem to reflect the
(Halmi et al. 1977; Goldberg et al. 1977; Garner desire to restrict food intake found in all patients
& Garfinkel, 1980). and the bulimia experienced by a specific
In addition, demographic and clinical features subgroup. The 14 items not loading on the 3
such as age of onset, duration of illness, factors were eliminated, leaving a new 26-item
percentage of average weight at testing, previous scale (EAT-26) which was included in the
minimum and maximum adult weights, and correlational analysis.
frequency of bulimic episodes were determined
for all patients. Comparison of EAT-26 and EAT-40
Table 2 indicates the mean differences between
the AN and FC groups on the EAT-40, EAT-26
RESULTS
and EAT-26 factor scores,1 as well as differences
Following Strober's (1981) caveat suggesting between bulimic and restricter subsamples of the
that, within a heterogeneous patient sample, age anorexia nervosa group. While bulimic and
and duration of illness may serve as confounding restricter patients did not differ in total EAT-40
variables, we performed a preliminary correla- or EAT-26 scores, the bulimics scored signifi-
tional analysis to determine the relationship cantly higher on Factor II and lower on Factor
between the EAT score and those clinical III than the restricters. Raw scores corresponding
features. None of the dependent or independent to percentiles for both the AN and FC groups are
variables in this study was significantly related to presented in Table 3. Since bulimic and restricter
the subjects' ages or the duration of illness in the subgroups differed significantly on Factors II
patient sample. and III, their norms are reported separately.
The EAT was analysed in two stages: first, a Table 3 also indicates the standardized
factor analysis was performed; this was followed reliability coefficients (Cronbach's alpha) for
by a correlational analysis of the total scale, both groups for factors, the EAT-26 and the
factor scores, clinical and other psychometric EAT-40. A 'cut-off score' of 30 has been
variables. established for the EAT-40 (Garner & Garfinkel,
1979). With the present sample, a cut-off score
Factor analysis of 20 on the EAT-26 correctly classifies a similar
The 40 items of the EAT were factor analysed for proportion of AN and FC subjects according to
the sample of 160 anorexia nervosa patients. group membership. Usingadiscriminant function
Three factors were extracted accounting for analysis, the overall percentage of cases correctly
40-2% of the total variance; the number of classified on the basis of total score was 84-9"0
factors derived was determined by the scree test, for the EAT-40 and 83-6% for the EAT-26.
described by Cattell (1966). An oblique rotation
was performed and the items with factor Correlational analysis of the EAT
loadings of 0-40 or above (in absolute value) are Table 4 presents the correlational analysis
reported in Table 1. Fourteen items did not load between the EAT (EAT-40, EAT-26. and
on any of the factors. Four items (numbers 4, 9, 1
Unit weights were used to culcuhuc factor scores.
874 D. M. Garner, M. P. Olmsted, Y. Bohr and P. E. Garfinkel

Table 1. EA T-40 factor structure

Item Corrected item Factor

number Item content total correlation loading

Factor 1: Dieting
37 Engage in dieting behaviour 0 71 0-72
30 Eat diet foods 0-61 0-69
36 Feel uncomfortable after 0-64 0-68
eating sweets
39 Enjoy trying new rich foods 0-44 0-66
29 Avoid foods with sugar in them 0-55 0-64
10 Particularly avoid foods with 0-54 0-58
high carbohydrate content
15 Am preoccupied with a desire 0-69 0-51
to be thinner
38 Like my stomach to be empty 0-62 0-48
22 Think about burning up calories 0-67 0-47
when 1 exercise
14 Feel extremely guilty after eating 0-68 0-46
4 Am terrified about being overweight 0-65 0-45
25 Am preoccupied with the thought 0-66 0-45
of having fat on my body
9 Aware of the calorie content 0-45 0-45
of foods that I eat
Factor II: Bulimia and food preoccupation
40 Have the impulse to vomit 0-59 0-78
after meals
13 Vomit after I have eaten 0-45 0-75
7 Have gone on eating binges 0-51 0-63
where 1 feel that I may not be
able to stop
34 Give too much time and thought to food 0-71 060
6 Find myself preoccupied with food 0-65 0-59
31 Feel that food controls my life 0-66 0-55
Factor III: Oral control
8 Cut my food into small pieces 0-62 0 81
26 Take longer than others to 059 0-69
eat meals
24 Other people think that I am 0-50 0-69
too thin
12 Feel that others would prefer 0-64 0-62
if 1 ate more
33 Feel that others pressure me to eat 0-64 0-62
5 Avoid eating when I am hungry 0-54 0-52
32 Display self-control around food 0-46 0-41

EAT-26 factors) and psychometric as well as is a better predictor of total EAT score and factor
clinical features within the AN sample. The total scores (except for Factor III) than the individual
EAT-40 and EAT-26 have similar significant body-image variables.
relationships with body-image variables and The intercorrelations between EAT variables
HSCL scores, but are unrelated to bulimia and suggest that the EAT-26 is highly predictive of
the percentage of average weight. Factor I is the total EAT-40 (r = 0-98) (Table 5). Moreover,
significantly related to body-image variables, but the EAT-26 maintains as robust a correlation
unrelated to bulimia. Factor II is similar to with clinical and psychometric variables as the
Factor I in its relationship to body-image vari- original scale, indicating that the 14 items
ables but, unlike Factor I, Factor II is strongly eliminated from the EAT-40 are redundant and
related to bulimia, a heavier body weight and do not increase the instrument's predictive
total symptoms on the HSCL. Factor III is capability. Factor I of the EAT-26 has the
negatively related to the percentage of average highest correlation with the total EAT-26
weight and to bulimia. It is not significantly (r = 0-93). Factors II (r = 0-64) and III
related to body-image variables or other psycho- (r = 0-60) have a weaker positive relationship to
metric tests. The Body-image composite score the EAT-26 total scale score, with Factor III
 The Eating Attitudes Test 875

Table 2. Mean EAT variable scores for anorexia nervosa (AN) versus female comparison (FC)
groups and bulimic (B) versus restricter (R) groups
AN (A1 =160) FC(A'=I4O)
Mean (s.D.) Mean (S.D.) ; value* P valuet

EAT-40 52-9 (230) 15 4 (110) 18-25 < 00001

EAT-26 361 (170) 9-9 (9-2) 16-90 < 00001
Factor I 19 9 (10-9) 7-1 (7-2) 1213 < 00001
Factor II 80 (5-2) 10 (21) 15-65 < 0-0001
Factor III 8-3 (5-8) 1-9 (21) 12-91 < 00001

AN (B) (AT = 83) AN (R) (AT = 77)

Mean (s.D.) Mean (S.D.) t value* P valuet

EAT-40 55 5 (21.0) 501 (24-9) 0-49 NS

EAT-26 38-4 (150) 33-7 (18 7) 0-46 NS
Factor I 20-8 (101) 18-8 (118) 0-42 NS
Factor II 10-6 (4-8) 5-2 (41) 7-78 < 00001
Factor III 70 (5-7) 9-7 (5-5) -312 < 0002

* Mest using separate variance estimate when group variance difference is significant (P < 005).
t Significance levels must be adjusted in consideration of the "error rate' for a family of multiple comparisons (Myers, 1979). If 0001 is
employed as the critical P value for individual comparisons, the error rate for each set of 5 comparisons is 0005.

Table 3. Norms: raw scores corresponding to percentiles for the EAT-26 and EAT-40 for anorexia
nervosa (AN) and female comparison (FC) groups
Factor I Factor II Factor III EAT-26 EAT-40
i—
Percentile AN FC AN FC AN FC AN FC AN FC

99 38 26 B*17\ B20 \
R 13 J
8
R 18/ 8 69 37 99 49

90 34 18 B 16 1 B 16 1
4 R 17 J 5 58 23 83 31
R 11 J
80 29 14 B 15 1 B 121
2 4 50 18 73 24
R9 / R 16 J
70 27 9 B 131 B 10 1
1 3 46 14 66 19
R8 / R 13 /
60 24 6 B 12 B8 1
R7 R 11 J 2 42 9 60 15
50 21 4 B 11 B5 I
1 36 6 52 11
R5 R 10 /
40 17 3 B 10 \ B4 32 5 46 9
0
R3 / R8
30 13 2 B8 B3 28 4 40 8
R 1 R7
20 8 B6 Bl |
R0 R5 ) 0 21 2 31 6

10 4 0 B3 B0 11 1 19 5
R R2
1 0 B0 B 1 0 6 0
R R0

Reliability 0-90 0-86 0-84 0-61 0-83 0-46 0-90 0-83 0-92 0-83
coefficient

*B, bulimic AN; R, restricter AN.

876 D. M. Gamer, M. P. Olmsted, Y. Bohr and P. E. Garfinkel

Table 4. Correlations between EAT variables, clinical features and psychometric measures for
anorexia nervosa
Variable (A-) Factor I Factor II Factor III EAT-26 EAT-40

% average weight (81) 0-22 0-54» -0-52* 013 006

Minimum weight (82) 0-31 0-51* -0-28 -0-26 -017
Maximum weight (82) -0-3 0-26 -014 002 002
Duration of illness (87) 0 12 001 001 009 012
(months)
Age (88) -001 -002 -012 003 008
Bulimia (frequency) (87) -008 0-60» -0-41* 009 004
Body size estimate (75) 0-50* 0-36* -010 0-42* 0-41*
Ideal size estimate (74) -0-47* -0-47* 0-25 -0-38* —0 41*
Body dissatisfaction (53) 0-41* 0-33 0-23 0-44* 0-45*
Body-image (50) 0-61* 0-48* 007 0-57* 0-56*
composite score
Locus of control (71) 019 0-29 -003 0-22 0-22
HSCL (Total) (54) 0-34 0-40* 0-21 0-42* 0-45*
Depression 0-32 0-37 018 0-38 0-41*
Anxiety 0-32 0-35 012 0-36 0-40
Interpersonal sensitivity 0-25 0-38 014 0-33 0-33
Obsessionality 0-28 0-27 014 0-31 0-34
Somatization 0-27 0-32 0-39 0-41* 0-45*

*P < 0001 (for individual comparisons).

Note. Significance levels have been adjusted in consideration of the increased possibility of Type I error for multiple comparisons. Since
95 individual correlations were computed, only individual P values of 0001 or less were considered to indicate significance. The corresponding
family error rate is 009. While this value is higher than conventional single comparison error rates, it does represent a compromise between
Type 1 and Type II error rates (Myers, 1979).

Table 5. Correlations between EAT-40, EAT-26 and EAT 26 factors for anorexia nervosa (AN)
and female comparison groups (FC)
AN(A>= 160)

EAT-40 EAT-26 Factor I Factor II Factor III

fEAT-40 0-98* 0-91* 0-59* 0-63*

EAT-26 0-97* 0-93* 0-64# 0-60'
FC
{ Factor I 0-94* 0-97* 0-51* 0-39'
( N = 140)1 Factorll 0-66* 0-67* 0-60* 002
[ Factor III 0-42* 0-44' 0-27* -003

• />< 0-001.

accounting for only 36% of the variance in the loading on the 3 factors. It was concluded that
EAT-26. the 14 items eliminated from the EAT-40 were
redundant, and did not increase the scale's
predictive power.
DISCUSSION The results from the factor analysis of the EAT
A previous report described 7 factors on the EAT must be viewed with caution, since it has been
(Garner & Garfinkel, 1979); however, in the recommended that the number of subjects
current study we chose to extract the smallest should be 10 times the number of items included
number of factors that was justifiable on in the analysis (Nunnally, 1967). Nevertheless,
statistical grounds. This factor analysis indicated the results from this tentative factor analysis
that the original 40-item scale may be abbreviated indicate that the EAT may be divided into
to provide a multi-factoral 26-item scale 3 distinct clusters of items. While there were
(EAT-26) which correlates highly with the no differences between bulimic and restricter
original scale (r = 0-98). Items were eliminated subtypes on total EAT score, these groups did
which did not load significantly on any of the 3 differ on 2 of the 3 EAT factors described in the
rotated factors and the score on the EAT-26 current study.
involves simply summing the scores for items The factor accounting for the largest amount
 The Eating Attitudes Test 877

of variance in the total EAT-26 was labelled on the EAT-26 could be a sensitive predictor of
'dieting' and reflects a pathological avoidance of outcome. Specifically, low scores on Factor II
fattening foods and shape preoccupations. While and high scores on Factor III may be better
factor I is associated with several parameters of predictors of favourable outcome than the total
the multidimensional body-image construct, it is scale score. However, as yet, outcome data are
not related to bulimia. AN subjects who score not available to test this hypothesis.
highly on Factor I may be described as over- Recent reports have indicated that cases of
estimators of their body size who are dissatisfied extreme dieting and bulimia are relatively
with their shape and desire to be smaller. Factor common among female university students
I is extremely reliable (alpha = 0-90) and is (Hawkins & Clement, 1980*; Wardle, 1980).
highly correlated with the total EAT-26 While these studies have suggested an analogy
(r = 0-93). Thus it could be used as an economical between the forms of behaviour observed in their
(13-item) substitute for the total scale in some samples and those observed in anorexia nervosa
circumstances. Factor II was labelled 'bulimia patients, their assessment measures have not
and food preoccupation' and, while similar to been standardized on subjects with eating
Factor I in its relationship to body-image distur- disorders.
bances, it differs in that it is positively related The EAT has been validated with anorexia
to bulimia and a heavier body weight. Since these nervosa patients but has also been useful in
characteristics have been associated with poor identifying eating disturbances in non-clinical
prognosis (Garfinkel et al. 1977; Slade & Russell, samples (Garner & Garfinkel, 1980; Button &
1973), it could be hypothesized that high scores Whitehouse, 1981; Thompson & Schwartz,
on Factor II would be associated with poor 1982). While most individuals from these
outcome. Factor III was labelled 'oral control' non-clinical groups who score highly on the EAT
and was largely comprised of items reflecting do not satisfy the diagnostic criteria for anorexia
self-control about food as well as those which nervosa, the majority have been identified (in
acknowledge social pressure to gain weight. High personal interviews) as experiencing abnormal
scores on Factor III are related to a lower weight eating patterns which interfere with normal
and the absence of bulimia. Again, because of the psychosocial functioning (Button & White-
previous findings of a negative relationship house, 1981; Garner & Garfinkel, 1979, 1980).
between prognosis and both self-overestimation Although the EAT may indicate the presence of
and bulimia (Garfinkel et al. 1977), it could be symptoms common to anorexia nervosa, as has
postulated that Factor III would be associated been cautioned previously, it would be inap-
with a favourable prognosis. Items on Factor III propriate to assume that high EAT scores are
acknowledge social forces in the environment diagnostic for anorexia nervosa in non-clinical
and thus high scores may indicate social aware- groups (Garner & Garfinkel, 1980; Button &
ness or responsiveness, which have been found Whitehouse, 1981). While the EAT may indicate
to reflect a good outcome (Crisp et al. 1979). the presence of disturbed eating patterns, it does
Norms have been presented for both the not reveal the motivation or possible psycho-
EAT-40 and the EAT-26 for a sample of pathology underlying the manifest behaviour.1
anorexia nervosa patients as well as a comparison Thus the EAT may be most suitable as either an
group of college females. The anorexic and outcome measure in clinical groups or as a
female comparison groups have significantly screening instrument in non-clinical settings.
different mean EAT-40 and EAT-26 total scores. In summary, the results from the current study
Both the EAT-40 and EAT-26 display acceptable have indicated associations between scores on
criterion-related validity by significantly predict- the EAT and various clinical and psychometric
ing group membership. The reliability (internal
consistency) of the EAT-26 is high (alpha = 0-90 1
We have recently reported on the development of an instrument,
for the AN group), even though this scale is much the Eating Disorder Inventory, which is designed to tap specific
psychological dimensions observed in anorexia nervosa and bulimia
shorter than the original EAT-40. (Garner et al. 1982). These psychological features include body-image
Because of the association with clinical and disturbances, ineffectiveness, interpersonal distrust, fear of psycho-
biological maturity and perfectionism, all of which have been identi-
psychometric features, it could be postulated fied as fundamental to anorexia nervosa (Bruch, 1973; Crisp, 1980;
that a particular configuration of factor scores Selvini-Palazzoli, 1974; and others).
878 D. M. Garner, M. P. Olmsted, Y. Bohr and P. E. Garfinkel

variables. A factor analysis of the EAT-40 has Garner, D. M. & Garfinkel, P. E. (1979). The Eating Attitudes Test:
an index of the symptoms of anorexia nervosa. Psychological
revealed item clusters which relate to specific Medicine 9, 273-279.
predictor variables. An abbreviated version of Garner, D. M. & Garfinkel, P. E. (1980). Socio-cultural factors in the
the original scale has been offered (EAT-26) development of anorexia nervosa. Psychological Medicine 10,
647-656.
based upon the items which load significantly on Garner, D. M. & Garfinkel, P. E. (1981). Body image in anorexia
the 3 factors. Whereas bulimic and restrictcr nervosa: Measurement, theory and clinical implications. Inter-
anorexia nervosa patients do not differ on the national Journal of Psychiatry in Medicine 11, 263 284.
Garner, D. M.. Garfinkel, P. E., Stancer, H. C. & Moldofsky. H.
total EAT-26 or EAT-40 scores, these groups do (1976). Body image disturbances in anorexia nervosa and obesity.
differ on Factors II and III of the EAT-26. The Psychosomatic Medicine 38, 227-237.
EAT-40 and EAT-26 are valid and economical Garner, D M . , Olmsted, M. P. & Polivy, J. (1982). The eating
disorder inventory: a measure of cognitive/behavioural dimensions
instruments which may be useful as objective of anorexia nervosa and bulimia. In Anorexia Nervosa: Recent
measures of the symptoms of anorexia nervosa. Developments (ed. P. L. Darby, P. E. Garfinkel, D. M. Garner and
D. V. Coscina). Allan R. Liss: New York (in the press).
The authors are grateful for the assistance of Ms M. Goldberg, S. C , Halmi, K. A., Casper, R., Eckert, E. & Davis, J. M.
O'Shaughnessy and Ms N. Resnick. This research was (1977). Pretreatment predictors of weight change in anorexia
supported by the Ontario Mental Health Foundation nervosa. In Anorexia Nervosa (ed. R. A. Vigersky), pp. 31—42.
Raven Press: New York.
(grant no. 810) and a Medical Research Council Goldberg, S. C , Halmi, K. A., Eckert, E. D., Casper, R. C , Davis,
Scholarship (to D.M.G.). J. M. & Roper, M. (1980). Attitudinal dimensions in anorexia
nervosa. Journal of Psychiatric Research 15, 239-251.
Halmi, K. A., Goldberg, S. C , Eckert, E., Casper, R. & Davis. J. M.
REFERENCES (1977). Pretreatment evaluation in anorexia nervosa. In Anorexia
Nervosa (ed. R. A. Vigersky), pp. 43-54. Raven Press: New York.
Berscheid, E., Walster, E. & Hohrnstedt, G. (1973). The happy Hawkins. R. C. & Clement, P. F. (1980a). Binge eating syndrome:
American body: A survey report. Psychology Today, Nov., 119- The measurement problem and a conceptual model. Presented at
131. the Annual Meeting of the Association for the Advancement of
Button, E. J. & Whitehouse, A. (1981). Subclinical anorexia nervosa. Behavior Therapy, New York.
Psychological Medicine 11, 509-516. Hawkins, R. C.&Clement.P. F. (1980 b). Development and construct
Bruch, H. (1973). Eating Disorders. Basic Books: New York. validation of a self-report measure of binge eating tendencies.
Cattell, R. B. (1966). The scree test for the number of factors. Addictive Behaviors 5, 219-226.
Multivariate Behavioral Research 1, 245-276. Health and Welfare Canada (1954). Canadian Average Weights for
Crisp, A. H. (1980). Anorexia Nervosa: Let Me Be. Grune and Height, Age and Sex. Nutrition Discussion of the Department of
Stratton: New York. Health and Welfare: Ottawa.
Crisp, A. H., Hsu, L. K. G. & Stonehill, E. (1979). Personality, body Hood, J., Moore, T. E. & Garner, D. M. (1982). Locus of control as
weight and ultimate outcome in anorexia nervosa. Journal of a measure of ineffectiveness in anorexia nervosa. Journal of
Clinical Psychiatry 40, 332-335. Consulting and Clinical Psychology SO, 3-13.
Derogatis, L., Lipman, R., Rickels, K., Uhlenhuth, E. H. & Covi, L. Nunnally. J. C. (1967). Psychometric Theory. McGraw-Hill: New
(1974). The Hopkins Symptom Checklist (HSCL): a self-report York.
symptom inventory. Behavioural Science 19, 1-15. Myers, J. L. (1979). Fundamentals of Experimental Design. Allyn and
Feighner, J. P., Robins, E., Guze, S. B., Woodruff, R. A., Winokur, Bacon Inc.: Boston.
G. & Munoz, R. (1972). Diagnostic criteria for use in psychiatric Reid, D. W. & Ware. E. E. (1973). Multidimensional^ of internal
research. Archives of General Psychiatry 26, 57-63. external control: implications for past and future research.
Fichter, M. M.& Keeser, W. (1980). Das Anorexia Nervosa inventar Canadian Journal of Behavioral Science 5, 264-271.
zur Selbstbeurteilung (AN1S). Archiv fur Psychiatric und Nerven- Selvini-Palazzoli, M. P. (1974). Anorexia Nervosa. Chaucer: London.
krankheilen 228, 67-89. Slade, P. D. (1973). A short anorexic behaviour scale. British Journal
Garfinkel, P. E. & Garner, D. M. (1982). Anorexia Nervosa: A of Psychiatry 122, 83-85.
Multidimensional Perspective. Brunner/Mazel: New York. Slade.P. D.&Russell,G. F. M.(I973). Awareness of body dimension
Garfinkel, P. E.. Moldofsky, H. & Garner, D. M. (1977). Prognosis in anorexia nervosa: cross-sectional and longitudinal studies.
in anorexia nervosa as influenced by clinical features, treatment and Psychological Medicine 3, 188-199.
self-perception. Canadian Medical Association Journal 117, Strober, M. (1981). The relation of personality characteristics to body
1041-1045. image disturbances in juvenile anorexia nervosa: a multivariate
Garfinkel.P. E., Moldofsky, H.& Garner, D. M.( 1979). The stability analysis. Psychosomatic Medicine 43, 323-330.
of perceptual disturbances in anorexia nervosa. Psychological Thompson, M. G. & Schwartz, D. M. (1982). Life adjustment of
Medicine 9, 703-708. women with anorexia nervosa and anorexic-like behavior. Inter-
Garfinkei, P. E., Moldofsky, H. & Garner, D. M. (1980). The hetero- national Journal of Eating Disorders 1, 47-60.
geneity of anorexia nervosa. Archives of General Psychiatry 37, Wardle, J. (1980). Dietary restraint and binge eating. Behavior
1036-1040. Analysis and Modification 4, 201- 209.