Journal of Psychosomatic Research 51 (2001) 757 – 764
Anger and personality in eating disorders
Secondo Fassino*, Giovanni Abbate Daga, Andrea Pierò, Paolo Leombruni,
Giovanni Giacomo Rovera
Department of Neuroscience, Psychiatric Institute, University of Turin, Via Cherasco 11, 10126 Torino, Italy
Received 2 June 2001; accepted 3 August 2001
Abstract
Objective: This study was designed to examine how anger,
temperament and character profiles differ across subtypes of eating
disorders (EDs) in comparison to healthy controls and to analyze
the relationship between anger expression, eating attitudes and
personality dimensions. Method: One hundred and thirty-five
outpatients (50 of whom suffered from anorexia nervosa restrictor
type [AN-R], 40 from anorexia nervosa binge/purging [AN-BP]
and 45 from bulimia nervosa [BN]) and 50 control subjects were
recruited and administered State-Trait Anger Expression Inventory
(STAXI), Temperament and Character Inventory (TCI) and Eating
Keywords: Eating disorders; Temperament; Character; Anger; Impulsivity
Introduction
Eating disorders (EDs) are severe illnesses characterised
by uncertain pathogenesis, early onset, long course and
significant therapeutic difficulties. Several psychiatric,
family and environmental stress factors can result in
EDs [1].
Clinical symptoms of EDs are various and complex; the
complexity of these disorders has led some authors to
study the personality features of persons suffering from
EDs, using two different approaches. The first is the
categorical approach and its objective is to diagnose
DSM-IV [2] Axis II Personality Disorders whose presence
might influence the course and outcome of EDs [3]. The
second is the dimensional approach whose objective is to
trace a basic personality profile expressing one’s risk of
developing an ED.
* Corresponding author. Tel.: +39-11-6634848, +39-11-6335425; fax:
+39-11-673473.
E-mail address: fassino@molinette.unito.it (S. Fassino).
0022-3999/01/$ – see front matter D 2001 Elsevier Science Inc. All rights reserved.
PII: S 0 0 2 2 - 3 9 9 9 ( 0 1 ) 0 0 2 8 0 - X
Disorder Inventory II (EDI-II). Results: STAXI showed greater
levels of anger in patients with BN than in those with AN. TCI
showed different personality profiles, in accordance with previous
studies. Correlations were found between the management of anger
feelings and psychological and personality traits typical of patients
with EDs. Conclusions: Clinically, impulsivity seems to be the
psychopathologic element most strongly correlated to anger.
Moreover, it appears clear that anger is better managed by
individuals with greater character strength. D 2001 Elsevier
Science Inc. All rights reserved.
The categorical approach has revealed a prevalence of
Cluster C personality disorders in 0 – 22% of anorectic
patients [4] and a prevalence of Cluster B personality
disorders in 2 – 50% of bulimic patients [4– 6].
Several authors have used the dimensional approach to
explore temperament and character dimensions of women
with EDs using the Tridimensional Personality Question-
naire (TPQ) and Temperament and Character Inventory
(TCI) developed by Cloninger and coworkers [7 –10]. This
approach allowed definition of some personality profiles
peculiar to each ED [11 –15].
Another research area attempts to identify the relevant
psychopathologic nuclei in EDs; this area has not yet
received much attention in the literature, but it might be
relevant for improvement in the treatment of EDs. The
identification of psychopathologic cores and of specific
temperament and character traits might be useful, at the
beginning of the treatment, to point out those patients with
greater risk to develop severe clinical symptoms of EDs.
This approach might provide course and outcome pre-
dictors, which at the moment are not sufficiently supplied by
Axes I and II diagnoses.
758 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764
Aggressiveness is a relevant psychopathologic core
because it can influence course and treatment outcome of
EDs [16 – 18]. The DSM-IV does not include a specific
aggressive disorder but considers only a very important
transnosographic dimension. Hostility, irritability and anger
are the most common expressions of aggressiveness. Low
tolerance and aggressiveness among individuals with EDs
are means of expression that, at least in part, may derive
from distorted family relationships and patterns [19,20] and
from childhood experiences [21]. Some authors have dem-
onstrated that in EDs correlation exists among severity of
disturbed eating patterns, low degrees of self-assertiveness,
high levels of self-directed hostility [22] and difficulty in
expressing anger [23]. Moreover, in these disorders, impul-
sive actions can be correlated with difficulty in expressing
anger [23].
Although some authors have investigated aggressiveness
and some affective states in EDs, data are still inadequate,
especially with regard to the relationship among anger, the
severity of the ED, and the basic personality of the
individual. Some peculiar behaviours of patients afflicted
with EDs, such as vomiting, might be associated with
higher levels of unexpressed anger, independently from
diagnosis [24].
In this work, anger has not been studied as a unitary
construct, but in its multifaceted nature, according to the
conceptualization of Spielberger [25]. This author has
stressed the importance of considering anger both as an
emotional state and as a trait. State-anger is a changeable
emotional condition, including feelings ranging from ten-
sion to fury, which are usually accompanied by symptoms
caused by the activation of the autonomic nervous system.
Trait-anger depends on the frequency of anger experiences,
defining the individual’s predisposition toward anger. More-
over, Spielberger stresses the fact that individuals are very
different in the way they suppress or express anger.
Thus, the aims of the current study were (1) to examine
the different expressions of anger in subjects with anorexia
nervosa restrictor type (AN-R), anorexia nervosa binge/
purging (AN-BP), bulimia nervosa (BN) and in a non-
clinical control group (CW); (2) to evaluate whether anger
expression is different in vomiting (AN-BP and BN) and
nonvomiting patients (AN-R); (3) to analyse the relation-
ship between anger expression and eating attitudes and
habits and (4) to analyse the relationship between anger
expression and the personality dimensions as evaluated
through the TCI.
Methods
Subjects
One hundred and thirty-five outpatients were recruited
from among the 218 who applied to the Eating Disorder
Pilot Center of the Psychiatric Clinic of the University of
Turin from October 1998 to August 2000. Fifty suffered
from AN-R (Group AN-R), 40 from binge/purging type
(Group AN-BP) and 45 from BN (Group BN). Fifty women
were recruited for the CW (Group CW).
On the basis of the inclusion criteria, outpatients
recruited in the study were women, aged 17– 32 years, with
a diagnosis of AN (restrictor or binge/purging type) or BN
according to DSM-IV criteria, but no Axis I comorbidity.
Thus, 21 patients with mood disorders, 5 with psychotic
disorders, 17 with anxiety disorders and 9 with any other
disorder pertaining to Axis I were excluded. This was done
to avoid the excessive heterogeneity of the sample in a
preliminary study and to increase the preciseness of the
diagnosis of ED.
The diagnosis and selection of outpatients was made by a
professional psychiatrist. Men, patients with a history of
EDs different from the ones currently being assessed, and
patients with medical conditions that were not homogeneous
to those of the rest of the group were excluded from the
study (m = 31). All patients were asked for informed written
consent to participate in the study.
The control group was recruited among a nonclinical
population of subjects matching those of the three clinical
groups in age and educational and socioeconomic levels.
Before participating in the study, each subject of the control
group was assessed by a psychiatrist to exclude any psy-
chiatric disorder.
Procedures
Diagnostic assessment for ED and other Axis I disorders
was carried out with the Structured Clinical Interview for
DSM-IV (SCID) [26]. A psychiatric expert performed a
screening interview (with SCID support) of approximately
an hour with every subject to determine possible inclusion
in the study. The following week, selected patients were
administered tests to evaluate ED psychopathology (Body
Mass Index [BMI], Eating Disorder Inventory II [EDI-II],
EDI Symptom Checklist [EDI-SC]) and personality and
anger (TCI, State-Trait Anger Expression Inventory
[STAXI]). None of the patients participating in the study
knew the meaning of these tests or had ever taken them.
None of the patients was treated with psychotropic drugs or
with psychotherapy at baseline.
Anger assessment
State-Trait Anger Expression Inventory (STAXI)
The STAXI [27] is a 44-item self-report questionnaire
that measures the experience and expression of anger; it
consists of 44 items that are divided into six scales and two
subscales. It measures the intensity of anger as an emotional
state (state-anger) and the disposition toward anger as a
personality trait (trait-anger). The Trait-anger scale contains
two subscales, T-Anger/T, which measures the general
disposition toward angry feelings (angry temperament),
 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764
and T-Anger/R, which measures the tendency to express
anger when one is criticized (reaction to criticism). Addi-
tional scales include Anger Expression-In, which measures
the frequency with which angry feelings are suppressed;
Anger Expression-Out, which measures the frequency of the
expression of anger toward other people or objects in the
environment and Anger Expression Control, which meas-
ures the frequency of attempting to control the expression of
anger. The final scale, AX/EX, gives a general index of the
expression of anger.
Participants rate themselves on four-point scales, assess-
ing either the intensity of their angry feelings or the frequency
with which anger is experienced, expressed, suppressed or
controlled [25]. In each case, higher scores indicate a greater
level of anger and its suppression or expression. The STAXI
has been validated on a variety of normal and clinical
populations and has good psychometric properties [27].
Personality assessment
Temperament and Character Inventory (TCI)
The TCI [10] is an inventory divided into seven inde-
pendent dimensions. Four of these test temperament (nov-
elty seeking [NS], harm avoidance [HA], reward
dependence [RD] and persistence [P]). Cloninger refers to
temperament as emotional responses that are moderately
heritable, stable throughout life and mediated by neuro-
transmitter functioning in the central nervous system and
that provide a clinical description of opposite extreme scores
[7,9]. Briefly, NS expresses the level of activation of
exploratory activity. Low NS corresponds with low explor-
ative activity, poor initiative, insecurity and unresponsive-
ness to novelty and change, whereas high scores express the
opposite characteristics. HA reflects the efficiency of the
behavioural inhibition system. Highly HA individuals are
described as extremely careful, passive and insecure, and
prone to react with a high rate of anxiety and depression to
stressful events. RD reflects the maintenance of rewarded
behaviour. Highly RD individuals are described as sen-
timental, attached, and easily influenced by others. P
expresses maintenance of behaviour as resistance to frustra-
tion. High P expresses the tendency to maintain unrewarded
behaviours and correlates with rigidity and obsessiveness.
The other three dimensions test character (self-directed-
ness [SD], cooperativeness [C], self-transcendence [ST]),
which are considered as personality traits acquired through
experience. SD expresses the degree to which the self is
viewed as autonomous and integrated. C reflects the degree
to which the self is viewed as a part of society. ST expresses
the degree to which the self is viewed as an integral part of
the universe. Low SD and C appear as a common denom-
inator extending across subtypes of personality disorders
and the most important predictor of categorical diagnosis of
a DSM Axis II disorder [10,28].
Each of the seven dimensions has several lower order
components; because they are not usually considered in the
759
literature, we will not report them here. The TCI displays a
good internal consistency (range 0.76 –0.89) [9].
Assessment of EDs
Body Mass Index (BMI)
The BMI, an index of body mass (weight/height2), is
related to the nutritional state of the subject. Female subjects
with a BMI between 18.7 and 23.8 are considered to be of a
normal weight [29].
Eating Disorder Inventory II (EDI-II)
The EDI-II [30,31] is a 91-item inventory that evaluates
the symptoms and the psychological characteristics of eating
behaviour disorders. It is composed of eight subscales and
three provisional subscales.
EDI Symptom Checklist (EDI-SC)
The EDI-SC [30,31] is a self-administered questionnaire
used to collect clinical data about ED: age of onset,
physical exercise, binges, purges, menses and use of
laxatives and diuretics.
Data analysis
The SAS was used for data analysis [32]. A one-way
analysis of variance (ANOVA) and the post hoc t test
(Duncan) were used to compare the means of four groups
(AN-R, AN-BP, BN and CW) on the age, BMI, EDI-II,
STAXI and TCI subscales. ANOVA and the post hoc t test
(Duncan) were used to compare the three groups for
duration of illness, and the t test for independent sample
for two groups regarding vomiting and frequency of vom-
iting. Since the four groups showed significant differences
in age, a general linear model (GLM) for ANOVA and for
ANCOVA was carried out for each scale, with the ‘‘age’’
variable taking part in the model as covariate. Then the four
groups were compared for all the scales with multivariate
analyses of variance and covariance (MANOVA and MAN-
COVA): all the significant variables to ANOVA and
ANCOVA were still significant. This confirms the inde-
pendence of the variables of the examined tests. The groups
were then compared with the post hoc t test (Duncan).
Corrective measures for the post hoc test (e.g., the
Bonferroni correction) were not used for two reasons: (1)
cogent arguments against the practice for exploratory stud-
ies have been put forward by the epidemiologist Rothman
[33] and (2) data dredging was avoided by conducting only
preplanned analysis [34].
The four groups were then reclassified into two groups
based on symptomotology: a GLM for two-way ANOVA was
carried out by using the variables diagnosis (AN-R, AN-BP,
BN and CW) and vomiting (yes or no) as control variables. A
level of significance of a
.05 was considered acceptable.
For the clinical groups the correlations among anger and
personality and anger and eating habits were evaluated
760 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764

Table 1
Anger differences between the three clinical groups and the control group: means ± S.D. of STAXI scores
Group
STAXI AN-Ra AN-BPb BNc CWd P* < P ** < Post-hocy
S-ANG 13.88 ± 5.59 13.72 ± 5.05 15.80 ± 6.60 12.72 ± 4.40 .054, F(2.60) .021, F(3.21) C > A, B, C
T-ANG 20.54 ± 5.43 20.02 ± 5.03 23.44 ± 5.91 19.38 ± 5.80 .003, F(4.70) .006, F(4.26) C > B, D
TANG/T 7.58 ± 2.62 7.57 ± 2.90 8.08 ± 2.96 6.64 ± 2.10 .062, F(2.48) .052, F(2.61)
T-ANG/R 9.34 ± 2.90 9.35 ± 2.68 11.37 ± 3.63 10.02 ± 2.58 .004, F(4.64) .007, F(4.11) C > A, B
AX/IN 19.36 ± 5.43 19.32 ± 4.60 19.73 ± 4.67 17.46 ± 4.76 .100, F(2.10) .110, F(2.10)
AX/OUT 15.62 ± 5.32 16.67 ± 5.18 18.22 ± 4.70 15.80 ± 3.90 .038, F(2.87) .022, F(3.17) C>D
AX/CON 20.20 ± 6.29 19.92 ± 9.46 18.17 ± 5.65 20.60 ± 6.07 .348, F(1.10) .287, F(1.28)
AX/EX 30.08 ± 12.77 32.52 ± 9.05 34.77 ± 9.12 28.38 ± 10.40 .021, F(3.30) .012, F(3.56) C>D
STAXI = State-Trait Anger Expression Inventory; S-ANG = intensity of anger as an emotional state; T-ANG = angry temperament; T-ANG/T = general
predisposition to feel or express anger without a specific reason; T-ANG/R = tendency to express anger provoked by criticism; AX/IN = suppressed anger; AX/
OUT = anger expressed toward other people or objects in the environment; AX/CON = anger control; AX/EX = general index of the expression of anger.
A P value < .05 was significant.
a
AN-R: anorexia restrictor.
b
AN-BP: anorexia nervosa binge/purging.
c
BN: bulimia nervosa.
d
CW: control women.
* P obtained with ANOVA.
** P obtained with ANCOVA controlling for age.
y
Significant post hoc comparisons with Duncan test ( P < .05).

using Pearson’s bivariate correlation method. The only
correlations that were taken into consideration were those
that registered r > .350 and P < .001. This was done to
reduce the risk of errors due to multiple comparisons.
Results
Demographics and clinical characteristics
The four groups were different in age, whose mean value
was 21.84 years (S.D. 5.87) in AN-R, 23.9 years (S.D. 4.99)
in AN-BP, 26.18 years (S.D. 5.79) in BN and 22.38 years
(S.D. 2.51) in CW. The four groups were also different in
BMI, with a mean value of 15.65 (S.D. 1.67) in AN-R,
16.17 (S.D. 1.22) in AN-BP, 21.50 (S.D. 3.25) in BN and
20.25 (S.D. 1.83) in CW. The duration of illness was
different in the three clinical groups, with a mean value of
44.9 months (S.D. 41.11) in AN-R, 75.27 (S.D. 54.52) in
AN-BP and 104.84 months (S.D. 54.38) in BN. The
frequency of vomiting episodes per week did not signifi-
cantly differ between the two groups that vomited (6.15
episodes/week, S.D. 7.62, for AN-BP; 6.81 episodes/week,
S.D. 8.47, for BN).

Table 2
Personality differences between the three clinical groups and the control group: means (S.D.) of TCI scores: means ± S.D. of TCI scores
Group
TCI AN-Ra AN-BPb BNc CWd P* < P ** < Post-hocy
NS 16.79 ± 6.10 19.77 ± 5.82 21.26 ± 6.55 18.50 ± 6.19 .005, F(4.42) .003, F(4.84) C>A
HA 22.79 ± 7.21 22.55 ± 6.49 21.28 ± 8.10 18.08 ± 7.03 .006, F(4.32) .006, F(4.24) D < A, B, C
RD 14.16 ± 4.14 15.47 ± 4.55 15.51 ± 4.56 16.26 ± 3.50 .097, F(2.13) .011, F(2.05)
P 4.61 ± 2.03 5.07 ± 2.99 3.95 ± 2.34 4.38 ± 1.92 .165, F(1.73) .092, F(2.18)
SD 22.55 ± 8.91 19.87 ± 6.71 20.04 ± 9.26 27.58 ± 8.55 .000, F(8.46) .000, F(7.57) D > A, B, C
C 29.53 ± 8.36 32.97 ± 3.51 28.28 ± 7.61 33.18 ± 4.95 .000, F(6.64) .000, F(7.39) D > A, C, B > C
ST 12.83 ± 6.28 15.85 ± 6.16 15.00 ± 5.42 13.98 ± 5.87 .096, F(2.14) .092, F(2.18)
TCI: temperamental traits: NS = novelty seeking; HA = harm avoidance; RD = reward dependence; P = persistence. Character traits: SD = self-directness;
C = cooperativeness; ST = self-transcendence.
A P value < .05 was significant.
a
AN-R: anorexia nervosa restrictor type.
b
AN-BP: anorexia nervosa binge/purging type.
c
BN: bulimia nervosa.
d
CW: control women.
* P obtained with ANOVA.
** P obtained with ANCOVA controlling for age.
y
Significant post hoc comparisons with Duncan test ( P < .05).
 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764 761

Table 3
Correlation among anger and EDI-II and TCI scores
S-ANG T-ANG T-ANG/T T-ANG/R AX-IN AX-OUT AX/CON AX-EX
EDI-II
DT .200* .253** .295** .199* .238** .179* .050 .173*
BU .112 .243** .057 .296** .192* .189* .035 .193*
BD .362** .220* .277** .165 .273** .071 .045 .190*
IN .445** .173* .256** .068 .439** .066 .020 .273**
Pe .168 .203* .077 .201* .263** .081 0.30 .102
ID .121 .154 .149 .160 .563** .026 .053 .191*
IA .268** .227** .174* .209* .491** .107 .098 .187*
MF .231** .199* .195* .098 .239** .041 .112 .209*
ASC .236** .286** .277** .146 .341** .134 .071 .198*
IR .406** .488** .528** .193* .287** .252** .130 .328**
SI .304** .211* .178* .144 .467** .073 .032 .277**

TCI
NS .141 .263** .221* .226** .174* .310** .331** .260**
HA .253** .191* .217* .012 .417** .160 .005 .264**
RD .074 .158 .177* .030 .146 .008 .021 .003
P .095 .062 .083 .103 .016 .122 .074 .069
SD .247** .463** .414** .316** .377** .389** .292** .520**
C .268** .521** .386** .472** .121 .439** .342** .452**
ST .098 .133 .075 .127 .046 .006 .020 .089
The strongest correlation is reported in boldface. EDI-II: DT=drive for thinness; BU=bulimia; BD=body dissatisfaction; IN=ineffectiveness; Pe=perfectionism;
ID=interpersonal distrust; IA=interoceptive awareness; MF=maturity fears; ASC=ascetism; IR=impulse regulation; SI=social insecurity.
For STAXI and TCI acronyms, see previous tables.
* P<.05.
** P<.001.

Anger Correlational analyses

Differences found (with one-way ANOVA and
ANCOVA) in STAXI scores among the four groups and
the results of post hoc comparison for each subscale are
reported in Table 1.
Temperament and character
Differences found in TCI (with one-way ANOVA and
ANCOVA) among the four groups and the results of
post hoc comparison for each subscale are reported in
Table 2.
Correlations were only discussed as significant when
P < .001 and r  .350. The correlations among STAXI sub-
scales and the EDI-II and TCI scores of the three clinical
groups are shown in Table 3. The correlations between EDI-
II and TCI results are not reported and discussed because
they are beyond the aims of the study.
Considering the three clinical groups, significant corre-
lations were not found among the scores of STAXI sub-
scales and clinic and demographic variables; age, BMI and
illness duration did not seem to be related to anger (STAXI)
in our sample (detailed data will be available on request for
interested readers).

Stratification of ED patients by vomiting status

Eighty (59.2%) of the 135 patients reported vomiting
(Vom+), whereas 55 (40.8%) did not (Vom  ). Significant
differences were observed (with a GLM for two-way
ANOVA, Diagnosis  Vomiting) between these two groups
only on the BU, F(9,98), P < .002, and IR, F(5,23),
P < .021, scales of EDI-II (these acronyms are defined in
the footnote to Table 3). For the scales that differed
significantly, scores were higher for those who vomited.
The means of these two groups did not differ on the STAXI
and TCI scores.
Discussion
Regarding the first aim of this study, the comparison
among the three clinical groups and the control group for
anger expression leads to discordant results.
The emerging anger profile is one of bulimic patients
whose temperament is more inclined to anger, especially as
a reaction to criticism; they also feel greater anger feelings
and express them toward other people or objects in the
environment. In fact, according to previous studies
[16,23,35] bulimic patients show higher anger levels than
762 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764
both the control group and the anorectic group. These
findings support the opinion that bulimic individuals have
a low tolerance to frustration and low impulse control,
which generate anger and irritability [36], whereas anorectic
subjects have pathologically high tolerance levels as a
consequence of conflict avoidance [37]. Actually, anorectic
patients tend to show poor self-assertiveness and to turn
hostility and anger toward themselves [22].
However, the two groups of women suffering from AN
have a similar profile with regard to anger management and
did not have higher levels of anger in comparison to the
control group, contrary to other studies [38]. This result
suggests three hypotheses. The first is that perhaps the
group of anorectic patients is a selected clinical population;
these subjects might be more disposed to treatment, less
opposed and thus less angry than other clinical anorectics.
The second is that the STAXI might not be effective in
discriminating the clinical group of anorectic patients from
the nonclinical control group on the basis of anger expres-
sion; for anger management, individuals with AN might be
more similar to the general female population than to
bulimic individuals. The third hypothesis is the one we
believe to be more likely based on our clinical experience
[39]; these individuals might express their anger even
through contradictory answers to the tests.
In addition to differences in anger, the TCI revealed
differences in temperament and character among the
groups. These findings corroborate those obtained by other
authors in previous studies [11 – 15] Both anorectic groups
showed higher levels of HA than the control group. As in
previous discussions [13,15], this supports the theory that
anorectic individuals tend to repress behaviours, suffer
from anticipation anxiety and find it difficult to adapt to
change [7]. The bulimic group had a higher NS than the
AN-R group, which is a typical trait among bulimic
individuals [11,14]. These data support the hypothesis that
this population is temperamentally predisposed toward
expressing angry feelings impulsively, excitability and
intolerantly [7].
For character differences revealed by the TCI, the three
groups of patients all had lower SD than the control group.
The AN-R and BN groups also showed less C than the
control group. According to Cloninger et al. [8,9] these two
dimensions are critically important in determining indi-
vidual strength of character and maturity level and the
potential risk of personality disorder, which is quite high
among the ED population [6,13,15], and perhaps also the
ability to cope with anger.
The second aim of this study was to evaluate whether
anger expression is different in vomiting (AN-BP and BN)
and nonvomiting subjects (AN-R and CW). This was
possible because anger is a transnosographic psychopatho-
logic aspect and the instruments used for the study of
personality (TCI) and of eating attitudes (EDI-II) are inde-
pendent from Axis I diagnosis and from the kind of sample
(clinical or not) [10,31].
The subjects who vomited did not show more anger than
those who did not. They had a more disturbed psychological
eating profile, especially in the area of impulsivity, accord-
ing to the EDI-II results.
We believe that these results, even if they are preliminary,
support the theory that among this vomiting population the
pattern of anger expression is similar to that of nonvomiting
eating disordered and of nonclinical controls. Instead, the
lack of the impulse control is peculiar, and is often related to
the difficulty in expressing angry feelings [37] or to char-
acter fragility [40]. Our results point out that impulsivity is
the only aspect that is not influenced by Axis I diagnosis.
Increasing the impulse control should thus be the first focus
of every therapeutic intervention for a vomiting patient.
Another objective of the study was to evaluate the hypo-
thesis that psychopathologic and personality features can
correlate with anger levels and with anger coping in individ-
uals with EDs. Aggressiveness and anger are common and
frequent among these patients and may be due to the duration
of illness and to unsuccessful therapy [41]. However, this
theory finds no support from our sample and the STAXI
scores do not coincide, demonstrating that this is a contro-
versial matter [41] that needs further study. No correlations
were found among anger, personality and BMI. It should be
stressed that, even if patients participating in the study were
seriously underweight, none of them had confounding med-
ical conditions at baseline (they were all outpatients).
The study of the correlations between anger and the eating
attitudes profile supports the theory that anger in the ED
population is correlated to impulsiveness [16,36,37]. Clin-
ically, impulsivity seems to be the psychopathologic element
that is most strongly correlated to anger. In fact, in our sample,
patients with higher levels of anger, both as an emotional state
(S-Anger) and as a trait (T-Anger), are also more impulsive.
Another emerging result is that patients who felt inad-
equate were more insecure and trusted less in interpersonal
relationships tended to repress angry feelings, in accord with
results of previous studies [22,42]. Specifically, anger sup-
pression correlates with interoceptive awareness. This last
result once more links EDs to psychosomatic disorders,
which have inhibition of emotional expression and particu-
larly a life-long tendency to suppress anger as common
features [43]. If we correlate the aspects associated with
anger to personality in EDs, it appears clear that anger is
better managed by individuals with greater character
strength [9]. Moreover, the tendency to suppress anger
correlates with high HA; this supports the fact that behav-
iour inhibition measured by HA on the TCI, which, accord-
ing to the authors, is often associated with depression [44],
is an index of a general difficulty in expressing feelings,
which is typical of patients with EDs [45].
Study limits
The use of self-reported questionnaires, the lack of data
collected by clinical interviews, the lack of data concerning
 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764
previous treatments and the exclusion of all patients with an
Axis I comorbidity are limitations that do not allow an
immediate generalization of the findings. On the other hand,
the sample size, the tests used and the preciseness of the
diagnosis of ED and of the subtypes are strengths of this study.
Conclusion
Thus, these findings require further investigation. Future
studies should evaluate how the different personality pro-
files and the correlated ways of expressing anger influence
the course of EDs and how they might be changed by actual
psychopharmacologic and psychotherapeutic treatments.
Acknowledgments
We thank Dr. C.M. Bulik for her comments on this report
and Dr. C. Gramaglia and S. Boggio for their help on this
work. The statistical analysis has been made by Dr. G.
Rocca, statistical consultant for medical research of
Direzione Sanitaria of ‘‘Molinette’’ Hospital, Turin, Italy.
References
[1] Gabbard GO, Atkinson SD, editors. Synopsis of treatments of psychi-
atric disorders. 2nd ed. Washington (DC): American Psychiatric
Press, 1996.
[2] American Psychiatric Association. DSM-IV. Diagnostic and statistical
manual of mental disorders. 4th ed. Washington (DC): APA, 1994.
[3] Halmi KA, Garfinkel PE. Eating disorders. In: Gabbard GO, Atkinson
SD, editors. Synopsis of treatments of psychiatric disorders. 2nd ed.
Washington (DC): American Psychiatric Press, 1996. pp. 875 – 935.
[4] Herzog DB, Keller MB, Lavor PW, Kenny GM, Sacks NR. The
prevalence of personality disorders in 210 women with eating disor-
ders. J Clin Psychiatry 1992;53:147 – 52.
[5] Wonderlich SA, Swift WJ, Slotnick HB, Goodman S. DSM-III-R
personality disorders in eating-disorder subtypes. Int J Eating Disord
1990;9:607 – 16.
[6] Matsunaga H, Kaye WH, McConaha C, Plotnicov K, Pollice C, Rau
R. Personality disorders among subjects recovered from eating disor-
ders. Int J Eating Disord 2000;27:353 – 7.
[7] Cloninger CR. A systematic method for clinical description and
classification of personality variants. Arch Gen Psychiatry 1987;44:
573 – 88.
[8] Cloninger CR, Przybeck TR, Svrakic DM. The Tridimensional Per-
sonality Questionnaire: US normative data. Psychol Rep 1991;69:
1047 – 57.
[9] Cloninger CR, Przybeck TR, Svrakic DM. A psychobiological model
of temperament and character. Arch Gen Psychiatry 1993;50:975 – 90.
[10] Cloninger CR, Przybeck TR, Svrakic DM, Wetzel RD. The Tempera-
ment and Character Inventory (TCI): a guide to its development and
use. St. Louis (MO): Center for Psychobiology of Personality, Wash-
ington University, 1994.
[11] Brewerton TD, Hand LD. The Tridimensional Personality Question-
naire in eating disorder patients. Int J Eating Disord 1993;14:213 – 8.
[12] Kleifield EI, Sunday S, Hurt S, Halmi KA. The Tridimensional Per-
sonality Questionnaire: an exploration of personality traits in eating
disorders. J Psychiatry Res 1994;28:413 – 23.
763
[13] Bulik CM, Sullivan PF, Weltzin TE, Kaye WH. Temperament in eat-
ing disorders. Int J Eating Disord 1995;17:251 – 61.
[14] Bulik CM, Sullivan PF, Joyce PR, Carter FA. Temperament, character
and personality disorder in bulimia nervosa. J Nerv Ment Dis
1995;185:704 – 7.
[15] Klump KL, Bulik CM, Pollice C, Halmi KA, Fichter MM, Berrettini
WH, Devlin B, Strober M, Kaplan A, Woodside DB, Treasure J,
Shabbout M, Lilenfeld LR, Plotnicov KH, Kaye WH. Temperament
and character in women with anorexia nervosa. J Nerv Ment Dis
2000;188:559 – 67.
[16] Fahy T, Eisler I. Impulsivity in eating disorders. Br J Psychiatry 1993;
162:193 – 7.
[17] Favaro A, Santonastaso P. Different types of self-injurious behavior in
bulimia nervosa. Compr Psychiatry 1999;40:57 – 60.
[18] Favaro A, Santonastaso P. Self-injurious behavior in anorexia nervosa.
J Nerv Ment Dis 2000;188:537 – 42.
[19] Shugar G, Krueger S. Aggressive family communication, weight gain,
and improved eating attitudes during systemic family therapy for
anorexia nervosa. Int J Eating Disord 1995;17:23 – 31.
[20] Fassino S, Svrakic D, Abbate Daga G, Amianto F, Leombruni P,
Stanic S. Anorectic family dynamics: temperament and character data.
Compr Psychiatry, in press.
[21] Steiger H, Goldstein C, Mongrain M, Van der Feen M. Description of
eating disorders, psychiatric and normal women along cognitive and
psychodynamic dimensions. Int J Eating Disord 1990;9:129 – 40.
[22] Millar HR, Williams GJ, Chamove AS. Eating disorders, perceived
control, assertiveness and hostility. Br J Clin Psychol 1990;29:327 – 35.
[23] Tiller J, Schmidt U, Ali S, Treasure J. Patterns of punitiveness in
women with eating disorders. Int J Eating Disord 1995;17:365 – 71.
[24] Milligan RJ, Waller G. Anger and bulimic psychopathology among
nonclinical women. Int J Eating Disord 2000;28:446 – 50.
[25] Spielberger CD. State-Trait Anger Expression Inventory: professional
manual. Odessa (FL): Psychological Assessment Resources, 1996.
[26] Spitzer RL, Williams JBW, Gibbon M. Instruction manual for the
structured clinical interview for DSM-III-R. New York: Biometrics
Research Department, New York State Psychiatric Institute, 1987.
[27] Spielberger CD. State-Trait Anger Expression Inventory. Firenze,
Italy: Organizzazioni Speciali, 1994.
[28] Svrakic DR, Whitehead C, Przybeck TR, Cloninger CR. Differential
diagnosis of personality disorders by the seven factor model of tem-
perament and character. Arch Gen Psychiatry 1993;50:991 – 9.
[29] Mitchell JE. Managing medical complications. In: Garner MG, Gar-
finkel PE, editors. Handbook of treatment for eating disorders. New
York: Guilford Press, 1997.
[30] Garner DM. Eating Disorder Inventory 2, 1984. Firenze, Italy: Organ-
izzazioni Speciali, 1995.
[31] Garner DM. Eating Disorder Inventory-2 professional manual. Odessa
(FL): Psychological Assessment Resources, 1993.
[32] SAS Institute. SAS/STAT Software: changes and enhancements for
release 6.12. Cary (NC): SAS Institute, 1996.
[33] Rothman KJ. Modern epidemiology. Boston: Little, Brown, 1986.
[34] Grove WM, Andreasen NC. Simultaneous tests of many hypotheses in
exploratory research. J Nerv Ment Dis 1982;170:3 – 8.
[35] Fava M, Rappe SM, West J, Herzog DB. Anger attacks in eating
disorders. Psychiatry Res 1995;28(56):205 – 12.
[36] Lacey JH. Self-damaging and addictive behaviour in bulimia nervosa.
A catchment area study. Br J Psychiatry 1993;163:190 – 4.
[37] Thompson KM, Wonderlich SA, Crosby RD, Mitchell JE. The neglect
link between eating disturbances and aggressive behavior in girls.
J Am Acad Child Adolesc Psychiatry 1999;38:1277 – 84.
[38] Engel K, Meier I. Clinical process studies on anxiety and aggressive-
ness affects in the inpatient therapy of anorexia nervosa. Psychother
Psychosom 1988;50:125 – 33.
[39] Fassino S, Abbate Daga G, Amianto F, Leombruni P, Fornas B, Gar-
zaro L, D’Ambrosio A, Rovera GG. Outcome predictors in anorectic
patients after 6 months of multimodal treatment. Psychother Psycho-
som 2001;70:201 – 8.
764 S. Fassino et al. / Journal of Psychosomatic Research 51 (2001) 757–764

[40] Bulik CM, Sullivan PF. Temperament, character and suicide attempts
in anorexia nervosa, bulimia nervosa and major depression. Acta
Psychiatr Scand 1999;100:27 – 32.
[41] Agras WS, Crow SJ, Halmi KA, Mitchell JE, Wilson GT, Kraemer
HC. Outcome predictors for the cognitive behavior treatment of buli-
mia nervosa: data from a multisite study. Am J Psychiatry 2000;157:
1302 – 8.
[42] Geller J, Cockell SJ, Goldner EM. Inhibited expression of negative
emotions and interpersonal orientation in anorexia nervosa. Int J Eat-
ing Disord 2000;28:8 – 19.
[43] Fava GA, Sonino N. Psychosomatic medicine: emerging trends and
perspectives. Psychother Psychosom 2000;69:184 – 97.
[44] Svrakic DM, Pryzbeck TR, Cloninger CR. Mood states and person-
ality traits. J Affective Disord 1992;24:217 – 26.
[45] Cochrane CE, Brewerton TD, Wilson DB, Hodges EL. Alexithymia in
eating disorders. Int J Eating Disord 1993;14:219 – 22.