This article was published in an Elsevier journal. The attached copy
is furnished to the author for non-commercial research and
education use, including for instruction at the author’s institution,
sharing with colleagues and providing to institution administration.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
Author's personal copy
Anxiety in the eating disorders: Understanding the overlap
Emma Pallistera, Glenn Wallerb,⁎
aSub-Department of Clinical Health Psychology, University College London, London, UK
bEating Disorders Service, Central and North West London NHS Foundation Trust and Eating Disorders Section,
Institute of Psychiatry, King's College London, London, UK
Received 20 December 2005; received in revised form 26 June 2007; accepted 8 July 2007
This paper reviews research investigating the comorbidity between eating disorders and anxiety disorders. Whilst there is some
inconsistency in the literature, it appears that women with eating disorders have higher rates of anxiety disorders than normal
controls. Potential causal relationships between eating disorders and anxiety disorders are outlined, though their relative
chronology appears to be somewhat inconsistent. Safety behaviours and cognitive avoidance strategies (i.e., cognitive narrowing
and blocking) are suggested as potential mechanisms linking the disorders. A model outlining this hypothesised relationship is
developed throughout the review. It is suggested that eating disorders and anxiety disorders might share common aetiological
factors, and that these factors can increase an individual's susceptibility to either disorder. Potential implications for the treatment
of eating disorders are outlined, and suggestions are made for further research.
© 2007 Elsevier Ltd. All rights reserved.
Cognitive model of anxiety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Phenomenological associations between anxiety and the eating disorders . . . . . . . . . . . . . . . . . . . . . .
2.1.Patterns of comorbidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.1.Comorbidity with generalised anxiety disorder . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2. Comorbidity with phobic anxiety. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.3. Comorbidity with agoraphobia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.4.Comorbidity with social phobia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.5. Comorbidity with obsessive-compulsive pathology. . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.6. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2. Temporal pattern. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mechanisms of association . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.Safety behaviours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.1.Cognitive and behavioural rigidity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.2.Illness-related behaviours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Available online at www.sciencedirect.com
Clinical Psychology Review 28 (2008) 366–386
⁎Corresponding author. Outpatient Eating Disorders Service, Vincent Square Clinic, Osbert Street, London SW1P 2QU, UK. Tel.: +44 20 8237
2104; fax: +44 20 8237 2280.
E-mail address: Glenn.Waller@iop.kcl.ac.uk (G. Waller).
0272-7358/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.
Author's personal copy
Cognitive avoidance strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
3.2.1. Cognitive narrowing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
3.2.2. Blocking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
3.2.3. Links with dissociation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
3.2.4. Conclusion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
Aetiological determinants of the eating-anxiety link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 379
Relationship to models of anxiety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
5.1. Beck and Clark's information processing model of anxiety . . . . . . . . . . . . . . . . . . . . . . . . . . 380
5.1.1. Attentional bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
5.1.2. Schema avoidance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
5.2. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
Final model of the association between anxiety and the eating disorders . . . . . . . . . . . . . . . . . . . . . . . 382
Implications for the treatment of the eating disorders . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
Suggestions for future research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
Body checking and avoidance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
Summary of the role of safety behaviours . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
It is difficult to establish accurate prevalence rates for eating disorders, both due to variations in methods used for
diagnosis and because much of the research has focused exclusively on the diagnoses of anorexia nervosa and bulimia
nervosa. Hoek (1993, 2002) describes how two-stage surveys (using screening measures followed by strict diagnostic
interviews) yield lower prevalence rates than earlier surveys based solely on questionnaires. On the basis on these
methods, he suggests that the average point prevalence rates amongst females1are just 0.28% for anorexia nervosa and
approximately 1.0% for bulimia nervosa. However, based on a range of epidemiological studies (including Hoek's,
1993 review), Fairburn and Harrison (2003) suggest slightly higher prevalence figures of 0.7% for anorexia nervosa,
and 1–2% for bulimia nervosa.Indeed, these rates are relatively consistent with the American Psychiatric Association's
prevalence estimates of 0.5–1% for anorexia nervosa, and 1–3% for bulimia nervosa (American Psychiatric
Of course, none of these prevalence figures takes into account the diagnosis of ‘eating disorder not otherwise
specified’ (EDNOS; DSM-IV, American Psychiatric Association, 1994), or the equivalent in ICD-10 of ‘atypical eating
disorders’ (World Health Organisation, 1992). These disorders appear to be at least as common in clinical practice as
anorexia nervosa and bulimia nervosa combined (Fairburn & Harrison, 2003; Fairburn & Walsh, 2002), and
consequently Fairburn and Walsh (2002) suggest that current prevalence rates underestimate the true number of eating
disorder cases. Indeed, Grilo (2002) cites evidence that binge eating disorder (one of the EDNOS disorders described in
DSM-IV), has prevalence rates of around 2–3% in community samples.
Despite the lack of clarity regarding their prevalence, it is clear that an eating disorder can impact negatively on the
sufferer's quality of life. Furthermore, eating disorders have been associated with a high rate of psychiatric
comorbidity. For example, Braun, Sunday, and Halmi (1994) reported that 81.9% of their sample of women with eating
disorders had at least one Axis I comorbidity, with depression, anxiety, and substance misuse being especially
common. Moreover, 69% of this sample met criteria for at least one personality disorder. Since the presence of
comorbidities is likely to complicate the formulation and treatment of eating disorders (O'Brien & Vincent, 2003),
there is a clinical need to understand more about their prevalence.
Of particular interest in this review are the different manifestations of anxiety found to be comorbid with eating
disorders. Although depression is the most frequently diagnosed comorbid disorder (e.g., Braun et al., 1994; Herzog,
Keller, Sacks, Yeh, & Lavori, 1992), eating disorder patients have reported that anxiety and anger are more likely to
drive binges than depression (Arnow, Kenardy, & Agras, 1992). In addition, anxiety has been associated with further
disordered eating behaviours, including vomiting (Carter & Duncan, 1984), laxative abuse (Weltzin, Bulik, McConaha,
& Kaye, 1995), and restriction (Chesler, 1995). Moreover, Weltzin et al. (1995) suggest that significant anxiety not
1Like the majority of eating disorders research, most prevalence studies are based on female samples, since only 5–10% of patients in clinical
samples are male (Hoek, 1993, 2002). The female pronoun will therefore be used throughout this review to refer to those with eating disorders.
367 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
only reduces compliance with therapy, but also often leads to premature termination of treatment. However, it remains
the case that the great majority of the evidence for anxiety as a risk factor in the eating disorders comes from
retrospective studies (e.g., Jacobi, Hayward, de Zwann, Kraemer, & Agras, 2004), and that almost all the evidence of
causal impact of emotional factors comes from experimental studies that have manipulated depression rather than
enxiety (e.g., Stice, 2002). Therefore, the potential for understanding the role of anxiety in the development and
maintenance of the eating disorders is still to be realised.
In this review of anxiety in the eating disorders, we begin by briefly outlining the general cognitive theory of
anxiety. We then discuss the patterns of comorbidity associated with eating disorders and anxiety disorders, and
delineate potential temporal relationships between these conditions. We go on to discuss how both safety behaviours
and cognitive avoidance strategies might be potential mechanisms linking anxiety and eating pathology, and consider
how cognitive avoidance strategies relate to information processing models of anxiety. To conclude the review, we
consider both potential implications for the treatment of eating disorders and avenues for future research.
1. Cognitive model of anxiety
Beck et al. have proposed a cognitive model of anxiety disorders (Beck, Emery, & Greenberg, 1985), which has
been adapted and elaborated by numerous authors, including Beck and Clark (1997). According to the Beck et al.
(1985) model, the central problem of anxiety involves the allocation of excessive threat meaning to innocuous
situations or stimuli, along with an underestimation of personal coping resources. They propose that this bias results in
cognitive, affective, physiological, and behavioural changes in the individual (Beck et al., 1985). Cognitive effects
might include fear- and vulnerability-related beliefs and images, difficulty concentrating and reasoning, and
hypervigilance for threat. These factors are likely to intensify the threat meaning assigned to the object or situation.
Physiologically, autonomic hyperarousal occurs in preparation for the fight, flight, freeze, or faint response, and
consequently the individual may be behaviourally mobilized (to fight or escape). However, this preparation for action
may be experienced as aversive physical symptoms (such as palpitations, breathlessness, or dizziness), which further
increase the individual's emotional arousal (e.g., Wells, 1997). An alternative physiological reaction suggested by Beck
et al. (1985) is that motor responses may be deactivated, resulting in immobilisation and feelings of helplessness and
increased vulnerability. In an attempt to prevent the anxiety response outlined above, anxious individuals inhibit risk-
taking and maximise safety-seeking and avoidance. However, as with escape, the use of safety behaviours and
avoidance strategies prevent the individual from learning that the imagined threat might be exaggerated (e.g., Wells,
1997). Hence, anxiety is maintained.
The above account gives a general description of the cognitive theory of anxiety. However, anxiety is a broad
concept incorporating a range of diagnoses, and of course the particular form anxiety takes may be different in different
diagnoses. Hence, not all anxiety disorders will be equally associated with eating disorders. The following section will
therefore consider the patterns of comorbidity found amongst eating disorders and a range of anxiety disorders.
2. Phenomenological associations between anxiety and the eating disorders
In this section, we consider phenomenological patterns of comorbidity between specific anxiety diagnoses and the
eating disorders, and the developmental sequencing of those patterns. However, much of the research cited in this
review is correlational in nature. It should therefore be remembered that whilst associations between disorders may
suggest possible aetiological links, causation cannot be inferred from correlation. In the following section, we consider
the causality behind those relationships.
2.1. Patterns of comorbidity
In order to understand the nature of associations more clearly, it is important to consider comorbidity from both
perspectives — the prevalence of anxiety disorders amongst people with eating disorders, and the prevalence of eating
disorders amongst people with anxiety disorders. However, given the relatively low prevalence of eating disorders
(American Psychiatric Association, 1994; Fairburn & Harrison, 2003; Hoek, 1993, 2002), there is less research
evidence regarding the latter. Even then, the limited evidence relates only to certain anxiety disorders. When reviewing
research on comorbidities, O'Brien and Vincent (2003) highlight the importance of considering factors such as sample
368 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
size and the presence or absence of comparison groups. These issues will therefore be highlighted in relation to the
studies presented below.
2.1.1. Comorbidity with generalised anxiety disorder
According to DSM-IV (American Psychiatric Association, 1994), generalised anxiety disorder (GAD) is
characterised by excessive anxiety and worry on a variety of topics, occurring on more days than not for at least six
months. Some studies have reported high rates of GAD amongst women with eating disorders. For example,
Schwalberg, Barlow, Algar, and Howard (1992) report lifetime prevalence rates of 55% in referred women with
bulimia nervosa, and 36.4% in obese binge eaters. However, the small sample sizes in this study (N=20 and N=22
respectively) are likely to affect the accuracy of the prevalence rates. In fact, as noted by Godart, Flament, Perdereau,
and Jeammet (2002), there are wide variations in the estimated prevalence rates of GAD across eating disorder studies.
For example, Lilenfeld, Kaye, Greeno, Merikangas, Plotnicov, Pollice et al. (1998) report a GAD lifetime prevalence
rate of just 13% amongst patients with bulimia nervosa, which was not significantly different to the prevalence rate
amongst controls (2%). On the other hand, they report prevalence rates of 31% amongst restrictive anorexics
(significantly different to the controls). However, in contrast, Garfinkel, Lin, Goering, Spegg, Goldbloom, Kennedy
et al. (1995) found that the lifetime prevalence rate of GAD amongst a community sample of women with bulimia
nervosa was significantly different to the rate amongst comparison women (10.9% and 2.5% respectively).
In their review of anxiety comorbidity in the eating disorders, Godart et al. (2002) highlight the heterogeneous
nature of the samples, methods, and results of previous research on GAD, and suggest that it is difficult to come to any
firm conclusions about prevalence rates. In response, Godart, Flament, Curt, Perdereau, Lang, Venisee et al. (2003)
designed a more controlled study, in which a larger sample size was used (271 women with a current eating disorder
diagnosis, and 271 matched controls from the community). They investigated a range of DSM-IV comorbid anxiety
disorders in this study, including GAD. Godart et al. (2003) reported significantly higher lifetime prevalence rates for
GAD amongst referred women with anorexia nervosa compared to controls (48.6% in the restricting subtype compared
to 3.6% amongst matched controls, and 45.4% in the binge/purge subtype compared to 10.9% amongst matched
controls). They also reported significantly higher rates of GAD amongst women with bulimia nervosa of the purging
subtype (32.6% compared to 7% amongst matched controls). Unfortunately, there were too few women with bulimia
nervosa of the non-purging subtype to detect differences in the rates of particular anxiety disorders.
Despite the fact that the above findings are not entirely consistent, GAD does appear to be a relatively common
comorbidity amongst eating disorder samples. Moreover, the results suggest that GAD is associated with the full range
of eating pathologies, and is not just associated with particular behaviours such as purging, bingeing, or restriction.
(although of course a diagnosis of GAD cannot be made unless there are also additional areas of worry).
2.1.2. Comorbidity with phobic anxiety
According to DSM-IV criteria, a specific phobia refers to marked and excessive fear of a particular object or
situation (e.g., heights, animals, receiving an injection). Exposure to this situation consistently provokes an immediate
anxiety response, which can take the form of a panic attack. As with generalised anxiety disorder, the literature on
comorbidity between eating disorders and simple phobia is somewhat inconsistent.
Bulik, Sullivan, Carter, and Joyce (1996) reported a 30% lifetime prevalence rate for simple phobias in their clinical
study of 114 women with bulimia nervosa. However, as there was no control group in this study, it is not clear how this
compares to normal controls. Wittchen and Essau (1993) suggest that simple phobia is the most frequent anxiety
disorder in community samples, affecting between 8.5% and 25.9% of women. Indeed, Lilenfeld et al. (1998) failed to
find a significant difference between prevalence rates in women with bulimia nervosa (19%) and control women (7%),
although they did find a significant difference between the control group and women with anorexia nervosa (27%).
Conversely, despite finding a difference in the overall prevalence of anxiety disorders, Halmi, Eckert, Marchi,
Sampugnaro, Apple, and Cohen (1991) found no significant difference in the lifetime prevalence of simple phobia in
women with a history of anorexia nervosa (followed up after ten years) compared to controls. However, in their
community study, Garfinkel et al. (1995) found significantly higher lifetime rates of simple phobia both in women with
full-syndrome bulimia nervosa (40.0%) and in those with partial-syndrome bulimia nervosa (40.9%) compared to
369 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
As with GAD, Godart et al. (2002) cite a wide range of prevalence figures, many based on small sample sizes.
Overall, the inconsistent results and range of prevalence rates cited (for both clinical and control participants) make it
difficult to draw any strong conclusions regarding the comorbidity of eating disorders and simple phobias. It is also not
possible to tell from the studies whether any particular phobias (e.g., emetiphobia) are more or less common in women
with eating disorders. It is of course possible that phobias related to eating (and associated situations) could be
subsumed under the more general heading of social phobia (see later section).
2.1.3. Comorbidity with agoraphobia
Agoraphobia is defined in DSM-IV as: “Anxiety about being in places or situations in which help may not be
available in the event of having an unexpected or situationally predisposed panic attack”. Halmi et al. (1991) report
significantly higher lifetime prevalence rates of agoraphobia in their follow up study of women with a history of
anorexia (most of whom had shifted diagnosis to EDNOS or bulimia nervosa, or were no longer diagnosable; 14.5%
compared to 3.2% amongst controls). Using a community sample, Garfinkel et al. (1995) also found a significantly
higher lifetime rate of agoraphobia amongst women with full-syndrome bulimia nervosa (34.5%) compared to controls
(7.5%). However, in their review of comorbidity amongst individuals with eating disorders, Godart et al. (2002) report
comorbidity rates for agoraphobia from 0% to 34.5%, and suggest that findings are “scarce and contradictory” (p. 267).
Following on from this, Godart et al. (2003) report significantly higher lifetime rates of agoraphobia in women with
anorexia nervosa compared to matched controls (19.8% in the restricting subtype compared to 2.7% amongst matched
controls, and 27.3% in the binge/purge subtype compared to 7.3% amongst matched controls). They also report
significantly higher lifetime rates of agoraphobia amongst women with bulimia nervosa of the purging subtype (17.4%
compared to 4.7% in matched controls). The non-purging bulimia group was too small to carry out statistical analyses.
Despite some mixed findings, the studies reviewed suggest that agoraphobia is more prevalent in eating disorder
samples compared to normal controls, although there is no apparent evidence that particular eating behaviours
(restricting, bingeing, or purging) are associated with this anxiety disorder.
2.1.4. Comorbidity with social phobia
Social phobia is characterised by a fear of social or performance situations owing to excessive concern over negative
evaluation by others (DSM IV; American Psychiatric Association, 1994). A number of studies have demonstrated
associations between social phobia and eating disorders. For example, Halmi et al. (1991) found significantly higher
lifetime rates of social phobia in their eating disorders group (33.9%) compared to controls (3.2%). Moreover, Godart
et al. (2003) observed similarly high lifetime and current rates of social phobia amongst their clients with eating
disorders, and reported that social phobia was the most frequent anxiety disorder in bulimia nervosa (lifetime
prevalence of 36.0% in the purging subtype compared to 8.1% amongst matched controls, and 36.8% in the non-
purging subtype; though this group was too small to detect differences with the controls). Social phobia was also the
second most frequent anxiety disorder in anorexia nervosa (lifetime prevalence rates of 37.8% in the restricting subtype
compared to 5.4% amongst matched controls, and 41.8% in the binge/purge subtype compared to 12.7% amongst
matched controls). Similar results have also been found in community based studies. For example, Garfinkel et al.
(1995) found significantly different prevalence rates between individuals with eating difficulties and those without
(45.5% in women with full-syndrome bulimia nervosa, 54.5% in women with partial-syndrome bulimia nervosa, and
15.2% amongst controls).
However, not all studies have reported this pattern of results. For example, whilst Bulik, Sullivan, Fear, and Joyce
(1997) found a significantly higher rate of social phobia amongst women with bulimia nervosa (30.2%) compared to
normal controls (6.1%), they did not find an increased rate amongst women with anorexia nervosa (5.9%). Bulik et al.
(1997) acknowledge that this latter finding is inconsistent with other comorbidity studies. By way of explanation, they
propose that despite an increased vulnerability, social phobia may not be expressed in women with anorexia nervosa
because situations that provoke social anxiety are typically avoided. Consequently, they suggest that the low
prevalence rate reported might reflect a lack of social interactions rather than a lack of social anxiety. It might also be
worth considering that the two eating disorder groups in this study were selected using quite different methods. Whilst
the bulimia nervosa group consisted of consecutive referrals to an outpatient treatment trial (for bulimia nervosa), the
anorexia nervosa group were identified and contacted on the basis of a retrospective review of clinical files (of
individuals referred to the eating disorders service between 1981 and 1984). Perhaps this group of women with
anorexia nervosa was therefore ‘atypical' in some way.
370E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
As mentioned previously, it can also be helpful to consider rates of eating disorders amongst samples of anxious
clients. For example, Wittchen, Stein, and Kessler (1999) investigated the prevalence of social phobia and related
comorbidities, using a community sample of 3021 individuals (male and female) between the ages of 14 and 24 years.
They found significantly higher rates of eating disorders (mostly EDNOS) among young people who met criteria for
generalised social phobia (15%) compared to those who did not (2.6%). However, since they do not specify the
EDNOS diagnoses given, it is not possible to consider whether particular kinds of eating pathologies are related.
Similarly, Becker, DeViva, and Zayfert (2004) investigated the rates of eating disorders amongst clients presenting at
an anxiety clinic. They found that nearly 12% of their sample met criteria for a probable eating disorder (based on a
questionnaire measure). Although the rates of probable eating disorder did not differ significantly across the anxiety
disorder diagnoses, social phobia had unique variance in the regression analysis.
Therefore, despite some inconsistencies, these studies suggest that social phobia is indeed more prevalent in eating
disorders, and that eating disorders might be more prevalent in social phobia cases. However, it is not clear from the
studies above how social phobia might be linked to different eating pathologies. However, Garfinkel, Lin, Goering,
Spegg, Goldbloom, Kennedy et al. (1996) found significantly higher lifetime rates of social phobia in individuals with
bulimia nervosa of the purging subtype (64.7%) compared to those with bulimia nervosa of the non-purging subtype
(31.1%). Furthermore, in their dimensional study of social anxiety (a necessary component of social phobia),
Hinrichsen, Wright, Waller, and Meyer (2003) investigated the levels of social anxiety across different eating disorder
diagnoses, as well as the types of emotion regulation mechanisms typically used. Based on questionnaire responses,
they found that social anxiety was higher amongst women with eating disorders (compared to non-clinical controls),
particularly those with binge-purge anorexia. Moreover, they found that higher levels of social anxiety were associated
with higher levels of bulimic pathology (as measured by questionnaire) in women with bulimia nervosa and non-
clinical controls. On the other hand, in the restrictive anorexic group, higher levels of social anxiety were associated
with higher levels of dissociation (again measured by questionnaire). However, despite finding the highest levels of
social anxiety amongst women with anorexia nervosa of the binge/purge subtype, this diagnosis was not associated
with any particular emotion regulation strategies.
2.1.5. Comorbidity with obsessive-compulsive pathology
The main feature of obsessive-compulsive disorder (OCD) is the occurrence of persistent and time-consuming
obsessions or compulsions (DSM IV; American Psychiatric Association, 1994). Obsessions may be inappropriate and
intrusive thoughts, impulses, or images (such as concern over contamination). Compulsions are repetitive behaviours
(either overt or covert), whose goal is to prevent or reduce anxiety or distress. In their review of obsessive-compulsive
disorder and anorexia nervosa, Serpell, Livingstone,Neiderman, and Lask (2002) conclude that a substantial number of
clients with anorexia nervosa display current obsessive and compulsive features, with an even greater proportion
meeting OCD criteria during their lifetimes.
Godart et al. (2002) report that OCD has been the most extensively studied anxiety disorder in eating disorder
samples, with significantly higher comorbidity rates consistently reported in clinical samples compared to controls. For
example, Halmi et al. (1991) report a lifetime prevalence rate of 25.8% amongst women with a history of anorexia
(followed up after ten years) compared to 6.5% amongst controls. Unfortunately, due to small sample sizes in this study
it was not possible to investigate whether lifetime rates of OCD are different amongst women with different current
eating pathologies. However, other research in this area suggests that increased OCD prevalence is more consistently
related to restrictive eating pathologies. For example, Godart et al. (2003) found significantly higher lifetime rates of
OCD in women with anorexia nervosa compared to controls (24.3% in the restrictive subtype compared to 2.7%
amongst matched controls, and 23.6% in the binge/purge subtype compared to 5.4% amongst matched controls).
However, they did not find significant differences between women with bulimia nervosa and controls. Whilst they
found OCD prevalence rates of 16.7% in women with non-purging bulimia nervosa, this was based on a small sample
size (N=19) and was therefore not significantly different to matched controls (5.3%). Among women with bulimia
nervosa of the purging subtype, the lifetime prevalence rate for OCD was 9.4%, which was again not significantly
different to matched controls (3.5%). Bulik et al. (1997) reported a similar pattern of results, finding a significantly
higher prevalence rate for OCD among women with anorexia nervosa (16.2%) compared to women with bulimia
nervosa (3.5%), and community controls (2%).
Despite finding a higher rate of eating disorders amongst anxious clients, Becker et al. (2004) did not find that OCD
contributed unique variance in their regression analysis. In light of this, they suggest that researchers should consider
371 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
that apparent relationships between OCD and eating disorders might be due to other comorbid anxiety disorders, rather
than being due to OCD per se. Moreover, it is important to consider that whilst comorbidity studies can suggest
associations between conditions, they cannot explain why that association exists. In fact, there are a variety of ways in
which comorbidities can be explained. These will be explored in the next section of this review.
In summary, it seems that OCD has been more clearly associated with restrictive eating disorder presentations, as
opposed to more bulimic presentations. However, the studies reviewed highlight the importance of considering the
existence of further comorbidities when considering the likely relationship between eating disorders and anxiety
Whilst the comorbidity literature is certainly not entirely consistent, the evidence reviewed suggests that women
with eating disorders do have higher rates of anxiety disorders compared to normal controls. In particular, higher rates
of generalised anxiety disorder, social phobia, and agoraphobia appear to be associated with all types of eating
pathology. On the other hand, obsessive-compulsive disorder has been more consistently associated with restrictive
Of course, the majority of studies reviewed were based on clinical samples, which means that comorbidity might be
overrepresented. Du Fort, Newman, and Bland (1993) suggest a number of potential reasons for this, including the
phenomenon known as Berkson's bias. This bias is seen in the fact that individuals with more than one psychiatric
illness are more likely to seek treatment than those with a single diagnosis. In addition, clinicians may be more likely to
refer individuals with multiple diagnoses (Feinstein, Walter, & Horowitz, 1986). However, whilst these factors might
limit the generalizability of clinical comorbidity studies to community samples, they are likely to give an accurate
representation of the levels of comorbidity typically seen in routine clinical practice.
2.2. Temporal pattern
Godart et al. (2003) highlight the importance of exploring the nature of the relationship between eating disorders and
anxiety disorders. They propose three possible explanations for the comorbidities found between these disorders,
suggesting that: (a) anxiety might be a risk factor for eating disorders; (b) anxiety might be secondary to an eating
disorder; or (c) the two kinds of disorders may have common shared vulnerability. Prospective studies of ‘at risk’
populations are difficult to carry out in this area, due to the low base rates of eating disorders (e.g., Serpell et al., 2002).
Consequently, research has typically focused on retrospective reports of the temporal relationship between anxiety and
eating disorders in order to draw inferences about possible aetiological links.
Where anxiety disorders are comorbid with eating disorders, anxiety is frequently reported to have developed prior
to the eating disorder. For example, Godart et al. (2003) found that of those women with comorbid anxiety problems,
47% had at least one anxiety disorder preceding the onset of their eating disorder. In particular, social phobia
commonly occurred first (between 51.2% and 71.4% of cases depending on the eating disorder subtype; with higher
rates associated with more bulimic pathologies). Similarly, using a large sample of adults with an eating disorder
diagnosis, Kaye, Bulik, Thornton, Barbarich, Masters, and Price Foundation Collaborative Group (2004) found that a
number of anxiety disorders (obsessive compulsive disorder – OCD; social phobia; specific phobias; generalised
anxiety disorder – GAD) typically began in childhood, and usually preceded the onset of the eating disorder. Other
anxiety-based disorders (post traumatic stress disorder; panic disorder; agoraphobia) tended to have their origins at
about the same age or later than the eating disorder. These findings were supported by a further study of comorbidity
with obsessive compulsive disorder (Kaye, Bulik, Thornton, Plotnicov, & Price Foundation Collaborative
Investigators, 2006). One potential explanation for these findings is that some forms of childhood anxiety have an
aetiological role in the development of eating disorders (as represented schematically in Fig. 1a). For example,
Schwalberg et al. (1992) suggest that anxiety about social evaluation might lead to excessive concerns over eating,
shape, and weight (all of which are known to be key elements in the pathology of eating disorders; e.g., Fairburn,
Cooper, & Shafran, 2003). Accordingly, social anxiety/phobia could be hypothesised to be a potential pathway to, or
risk factor for, eating disorders.
In relation to restrictive pathologies in particular, Bulik et al. (1997) found that where OCD was comorbid with
anorexia nervosa, most individuals reported that the OCD occurred prior to the onset of the eating disorder. Indeed,
Bulik et al. (1997) propose that OCD might increase the probability that an eating disorder is expressed as anorexia
372 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
nervosa (though they do not indicate how this effect might occur). Related to this question, Jacobi et al. (2004) suggest
that the interaction between anxiety and specific prodromal symptoms of eating disorders needs to be further explored.
Of course, anxiety disorders do not invariably precede the onset of eating disorders. In contrast to the findings
discussed above, Godart et al. (2003) found that OCD, along with panic disorder and GAD, more typically appeared
simultaneously or after the onset of the eating disorder. Kaye et al. (2004) similarly found that panic disorder and
agoraphobia more usually developed subsequently. This leaves the possibility that eating disorders could produce or at
least exacerbate anxiety (including OCD and GAD symptoms) in some individuals. This potential relationship between
eating disorders and anxiety is represented in Fig. 1b.
Part of this association might be connected to starvation effects. For example, Keys, Brozek, Henschel, Mickelsen,
and Taylor (1950) demonstrated that caloric deprivation could cause obsessive-compulsive features in previously
asymptomatic individuals. They found that following a prolonged period of food restriction; obsessions and
compulsions concerned with the preparation and consumption of food were particularly common. Serpell et al. (2002)
suggest that this study provides evidence for the possibility that some of the OCD symptoms observed in eating
disorders (particularly anorexia nervosa) might result from malnourishment, rather than being an indication of a ‘true’
OCD disorder. Pollice, Kaye, Greeno, and Weltzin (1997) also suggest that weight status can impact on general anxiety
symptoms. Theyinvestigated women with anorexia at different stages of illness and recovery. They found that although
anxiety and obsessionality scores remained significantly elevated in women whose weight had been restored
(compared to healthy controls), these scores were highest in women who remained underweight.
The evidence reviewed above suggests that the relative chronology between eating disorders and anxiety disorders
can vary depending on the anxiety disorder in question. Whilst some anxiety disorders (notably social phobia)
frequently precede the onset of the eating disorder, others more commonly occur simultaneously or later (for example
GAD). It also appears that whilst OCD features might exist prior to the onset of the eating disorder, they can also occur
subsequent to the eating disorder onset (perhaps as a result of malnutrition). However, in interpreting these temporal
patterns, it is important to take into account the average age at onset for the disorders (Godart et al., 2002, 2003). For
example, whilst social phobia typically starts in childhood, generalised anxiety disorder starts on average at 20 years of
age (American Psychiatric Association, 1994). Consequently, Godart et al. (2002, 2003) state that it is unclear whether
the relative chronology of anxiety disorders and eating disorders is causative, or whether it merely derives from the
natural course of both disorders. Indeed, these average ages at onset are consistent with the finding that social phobia
Fig. 1. Possible phenomenological links between anxiety and eating disorders, showing different possible models of risk for comorbidity.
373 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
frequently precedes the eating disorder, whereas generalised eating disorder more typically develops simultaneously or
after the eating disorder (Godart et al., 2003).
An alternative explanation for the frequent comorbidity between anxiety and eating disorders is that rather than one
causing the other; they both result from a common vulnerability factor (Godart et al., 2003; Wonderlich & Mitchell,
1997). For example, it could be hypothesised that early experiences (e.g., childhood teasing and bullying) could be a
risk factor for both disorders (see Fig. 1c). Family studies and multivariate twin studies are two methodologies that can
potentially be used to evaluate common cause models such as this (e.g., Lilenfeld, Wonderlich, Riso, Crosby, &
Mitchell, 2006). These methodologies will be described further in relation to Fig. 2.
Of course, the three models outlined in Fig. 1 are quite simplistic and suggest that causation is unidirectional. In
reality, the connection between eating disorders and anxiety disorders is likely to be more complex and bidirectional.
For example, in their transdiagnostic model of the eating disorders, Fairburn et al. (2003) describe how some clients
with eating disorders are unable to cope appropriately with intense emotional states such as anxiety. As a result, they
engage in “mood modulatory behaviour” (p.517), such as alcohol or drug use, self-harm, or dysfunctional eating
behaviours (such as bingeing). However, the persistent use of these behaviours is likely to result in increased negative
affect, which further reinforces their use (i.e., according to this model there is likely to be a bidirectional relationship
between negative affect and eating behaviours).
Furthermore, Godart et al. (2003) raise the possibility that the relationship between anxiety and eating disorders
might be mediated by a mutual relationship with depression (i.e., that depression is responsible for all or some of the
association between anxiety and eating disorders). Indeed, Jordan et al.(2003) found that women with anorexia nervosa
who were currently depressed were significantly more likely to have social phobia than those without comorbid
Although a detailed investigation of mediator effects is beyond the scope of this review, O'Brien and Vincent (2003)
suggest that future research on comorbidities in the eating disorders should involve the explicit testing of both mediator
and moderator relationships (the latter being a variable that influences the predictive power of an independent variable).
They propose that an understanding of the nature of these relationships has important implications for determining the
appropriate focus of treatment. For a thorough description on how to investigate and analyse moderator and mediator
effects using multiple regression, see Frazier, Tix, and Barron (2004).
Whilst we have outlined the possible temporal relationships between eating disorders and anxiety disorders, the
available evidence does not allow clear causal connections to be made. One of the factors preventing this is the absence
of prospective studies, which are difficult to undertake due to the low base rates of eating disorders (Serpell et al.,
2002). However, despite the uncertainty about any aetiological connections, the research evidence does suggest that the
eating disorders are associated with higher rates of anxiety disorders. It is therefore possible to discuss potential
mechanisms linking anxiety and eating pathology in individuals who experience both. Such mechanisms will be
outlined in the next section of this review.
Fig. 2. Safety behaviours as a mechanism of association between anxiety and eating disorders.
374 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
3. Mechanisms of association
According to cognitive models of anxiety (e.g., Beck & Clark, 1997; Beck et al., 1985), anxiety arises when a
situation is viewed as dangerous in some way. Individuals are more susceptible to appraising situations as threatening
if they have underlying schemas that represent the world as unsafe and themselves as vulnerable (Beck et al., 1985).
The anxiety response itself (which has an effect at the cognitive, affective, physiological, and behavioural levels) can
reinforce an individual's view of herself as vulnerable. For example, it may have a negative impact on their
functioning, or else be interpreted as a sign of serious physical (or mental) disorder. Furthermore, individuals may
inadvertently reinforce their anxiety through the use of safety behaviours and avoidance strategies, which prevent
disconfirmation of their belief in catastrophe/vulnerability and may even exacerbate their anxiety symptoms (e.g.,
A number of behaviours that are commonly associated with eating disorders can also be characterised in terms of
safety seeking (e.g., extreme restriction to avoid the fear of massive weight gain). In addition, cognitive avoidance
strategies (designed to avoid aversive self awareness and the associated negative affect) have been theoretically linked
to dysfunctional eating behaviours. The following sections will therefore describe safety behaviours and cognitive
avoidance strategies, and discuss how each could be a mechanism of association between anxiety and eating disorders.
3.1. Safety behaviours
Safety behaviours are behaviours that are used by anxious individuals as a way of gaining control and preventing a
feared catastrophe (Salkovskis,1991). However, whilst they may reduce anxiety in the short term, they help to maintain
it in the long term. This is because the non-occurrence of the feared outcome is attributed to the use of the safety
behaviour, rather than the improbability of the catastrophe (Salkovskis, 1991). Given the high rates of comorbidity
between eating disorders and anxiety disorders (described above), women with eating disorders may potentially use a
number of non-specific safety behaviours. However, there might also be some safety behaviours that are more
particular to the eating disorders, and to the characteristic concerns with loss of control over eating, shape, and weight
(e.g., Fairburn & Harrison, 2003). The following sections will therefore outline how cognitive and behavioural rigidity,
illness-related behaviours, and body checking might all be examples of safety behaviours in this population.
3.1.1. Cognitive and behavioural rigidity
Women with eating disorders frequently report rigid eating patterns, often with quite ritualistic elements. This may
include eating slowly, eating foods in a particular order, or eating only ‘safe’ foods (with little or no variation).
Modifying these eating patterns is often a challenge in therapy, as individuals can be afraid that eating outside of these
constraints will result in catastrophic outcomes, such as uncontrolled and rapid weight gain. Rigid daily eating patterns
can be evident both in primarily restrictive eaters and in those with bulimic behaviours. In the latter, the individual's
intended eating plan may particularly emphasise the need to avoid eating ‘forbidden’ foods, which are typically foods
that are perceived as risky and fattening (e.g., Fairburn, Cooper, & Cooper, 1986). However, it has been hypothesised
that attempted abstinence from ‘forbidden’ foods actually contributes to bingeing. For example, Kales (1990) suggests
that when women with eating disorders eat even a small amount of these foods, they often conclude that control over
eating has been lost, and hence binge eating ensues. Indeed, using a sample of women with bulimia nervosa, he found
that 69% of binges contained foods that these women defined as “forbidden,” whereas only 15% of normal eating
Of course, rigid and obsessional characteristics in the eating disorders do not always represent safety behaviours.
For example, in low weight individuals, rigidity may be a symptom of malnutrition and semi-starvation (Keys et al.,
1950; Serpell et al., 2002). However, where rigid behaviours appear to be related to the prevention of a feared
catastrophe, then (as in the treatment of anxiety disorders) clients can be encouraged to test out their beliefs by dropping
their safety behaviours (e.g., Wells, 1997).
3.1.2. Illness-related behaviours
Gastrointestinal symptoms (such as irritable bowel syndrome and dyspepsia) are common in both anorexia nervosa
(e.g., Waldholtz & Andersen, 1990) and bulimia nervosa (e.g., Chami, Andersen, Crowell, Schuster, & Whitehead,
1995). These symptoms are typically considered to be physiological consequences of disordered eating behaviours,
375 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
such as restriction, vomiting, and laxative misuse (e.g., Beumont, Russell, & Touyz, 1993). However, as with cognitive
and behavioural rigidity, the concept of safety-seeking might be relevant to these behaviours. Bethune, Gompels, and
Spickett (1999) cite three case studies in which individuals undertook strict low carbohydrate diets in response to a
perceived carbohydrate allergy. They suggest that the most likely explanation for the symptoms was in fact abnormal
glucose metabolism secondary to the carbohydrate restriction itself. The avoidance of carbohydrates might therefore be
hypothesised to be a form of safety behaviour, designed to prevent a feared consequence (gastrointestinal discomfort).
However, as with safetybehaviours generally (e.g., Salkovskis, 1991), the low carbohydrate diet apparently maintained
and even exacerbated the symptoms experienced. Of course, whilst Bethune et al.s' (1999) study is consistent with this
explanation, it does not examine it explicitly. Consequently further research (examining the thoughts and beliefs
driving these behaviours) is needed. Nonetheless, where food intolerances and allergies are reported, it may be
pertinent for clinicians to consider whether some form of safety behaviour is involved in their aetiology and
maintenance. Again, where this applies, the most suitable treatment is likely to involve testing out the consequences of
dropping the identified safety behaviour/s.
3.1.3. Body checking and avoidance
Body checking is another potential safety behaviour that is known to be common amongst women with eating
disorders (e.g., Cooper, Todd, & Wells, 2000; Shafran, Fairburn, Robinson, & Lask, 2004). Fairburn et al. (2003)
suggest that body checking stems from the over-evaluation of eating, shape, and weight typically seen in the eating
disorders. Individuals who engage in this process examine their body using a variety of methods, including repeated
weighing, scrutinizing specific body parts (with or without the aid of a mirror), judging their shape and weight using the
fit of clothes, or seeking reassurance from others (Rosen, 1997). Mountford, Haase, and Waller (2006) suggest that a
range of beliefs drive such body checking. They found evidence of beliefs that body checking helps the individual
maintain control over their eating and weight, beliefs that checking directly decreases anxiety (as a form of
reassurance), and beliefs that not checking results in a feared catastrophe (as is typical with safety behaviours).
Consequently, Mountford et al. (2006) suggest that it is simplistic to suggest that there is only one explanation for
checking behaviours, though it appears that at times it functions as a safety behaviour.
In addition to body checking, individuals may engage in avoidance behaviours. Shafran et al. (2004) found that
many individuals alternate between body checking and avoidance, the latter of which involves refusing to be weighed,
covering mirrors, and wearing baggy clothes to disguise their shape. As with checking behaviours, avoidance may be
related to a range of beliefs, including those related to averting a feared catastrophe. However, where they function as
safety behaviours, both checking and avoidance are likely to prevent individuals from disconfirming their fears (e.g.,
Salkovskis, 1991). This means that anxiety (including that related to eating, shape, and weight concerns) is likely to be
maintained in the long term.
3.1.4. Summary of the role of safety behaviours
Within the cognitive-behavioural framework for anxiety, individuals with underlying vulnerability cognitions are
seen as more likely to appraise situations as threatening. Consequently, they are more likely to use safety behaviours
and avoidance strategies. A number of behaviours commonly seen in the eating disorders can also be characterised in
terms of safety seeking. In this review, cognitive and behavioural rigidity, illness related behaviours, and body checking
(and avoidance) have all been highlighted as potential examples of safety behaviours. Whilst these behaviours might be
driven by a variety of beliefs, it appears that averting feared catastrophes (a characteristic of safety behaviours) is one of
their possible roles. It has therefore been suggested that safety behaviours (driven by underlying vulnerability
cognitions) might be a potential mechanism of association between anxiety and eating disorders. This is illustrated in
The model depicted in Fig. 2 suggests that underlying vulnerability cognitions result in cognitions relating to harm
avoidance. These are the cognitions that drive safety behaviours. However, it is suggested that triggers in the
environment determine the way in which these behaviours manifest. Given one set of environmental circumstances, the
cognitions may manifest as disordered eating behaviours, but given a different set, they may manifest as part of an
anxiety disorder. Safety behaviours are also seen as maintaining the disorder (eating disorder, anxiety disorder, or both)
because the non-occurrence of the feared outcome is attributed to the use of the safety behaviour rather than the
improbability of the feared event. This would lead to vulnerability cognitions being reinforced and maintained, leading
to further use of safety behaviours.
376E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
3.2. Cognitive avoidance strategies
Cognitive avoidance strategies are another potential mechanism for the association between anxiety disorders and
eating disorders. These strategies function either directly or indirectly to reduce awareness of negative cognitions and
the associated negative affect (e.g., anxiety or depression). Two main avoidance strategies have been described in the
literature: ‘cognitive narrowing’ and ‘blocking’. These strategies have been associated with a number of impulsive
behaviours, including self-harm (e.g., Suyemoto, 1998) and substance misuse (e.g., Steele & Josephs, 1988). However,
in relation to the eating disorders, they have been particularly associated with bingeing and vomiting. In the following
section, we begin by outlining Heatherton and Baumeister's (1991) escape from awareness model, which relates the
process of cognitive narrowing to binge eating. We then go on to describe the blocking models of Lacey (1986) and
Root and Fallon (1989), and discuss how this strategy might be relevant to bulimic behaviours. The construct of
dissociation will then be discussed in relation to these cognitive strategies.
It should be noted that because the eating disorder models for cognitive avoidance have been explicitly developed to
account for binge eating, they do not relate to restrictive attitudes and behaviours. However, it may be that synonymous
processes are associated with restrictive eating pathologies. This is therefore potentially an area that would benefit from
further research and theory development.
3.2.1. Cognitive narrowing
According to Heatherton and Baumeister's (1991) escape from awareness model, bingeing occurs as a consequence
of ‘cognitive narrowing’, an intentional strategy employed to escape from aversive self-awareness and the associated
negative affect. In support of this model, Heatherton and Baumeister (1991) cite evidence that bulimics have
unrealistically high standards or ideals, both in relation to bodily thinness (e.g., Powers, Schulman, Gleghorn, &
Prange, 1987), and in relation to performance and achievement (e.g., Butterfield & Leclair, 1988). Heatherton and
Baumeister (1991) suggest that these women experience aversive self-awareness when they perceive that they have
failed to live up to their demanding ideals. In order to escape from the associated emotional distress (e.g., anxiety and
depression), it is hypothesised that the cognitive strategy of ‘narrowing’ is employed. This strategy involves focusing
attention on the immediate and present stimulus environment and engaging in low-level or concrete thinking.
Consequently, self-awareness remains at a relatively low level, and negative perceptions of the self are not processed in
any depth (so negative emotions can be temporarily avoided). However, a further result of this shift to low levels of
thinking is that higher-level cognitive functions (such as reasoning and inhibition) are lost. To illustrate the
hypothesised effect on inhibitions, Heatherton and Baumeister give the example of cheating in an exam. They suggest
that at a low level of meaning, cheating can be reduced to making marks on a page (a far more tolerable act).
Accordingly, internal obstacles to this action are removed, making the individual more willing to carry it out. In a
similar way, cognitive narrowing is said to result in restraints against eating being reduced or removed from awareness,
so that bingeing occurs despite the binge-eater's abhorrence of the behaviour.
Indirect support for the connection between binge eating and reduced self-awareness comes from the finding that
instructions to monitor eating appear to enhance dietary restraint and reduce the likelihood of disinhibited eating
amongst dieters (e.g., Pecsok & Fremouw, 1988). In addition, alcohol consumption (which is known to reduce self-
awareness and behavioural inhibition) often precedes bingeing (e.g., Abraham & Beumont, 1982). There is also some
evidence to support the hypothesised change to lower-level thinking accompanying cognitive narrowing. For example,
dichotomous cognitions have been found to be prominent among both bulimics and subclinical binge eaters prior to
binge episodes (e.g., Lingswiler, Crowther, & Stephens, 1989). However, Heatherton and Baumeister (1991)
acknowledge that there is some ambiguity in the evidence. Indeed, while binge eating has often been associated with
reductions in anxiety (e.g., Fairburn et al., 1986), Heatherton and Baumeister suggest that it is not clear whether binge
eating results from this cognitive narrowing, or whether bingeing is itself able to facilitate the avoidance of negative
cognitions and emotions. The latter would perhaps be more consistent with the ‘blocking’ models described by Lacey
(1986) and by Root and Fallon (1989).
In contrast to the cognitive narrowing model, blocking models depict binge eating as a direct method for reducing
aversive self-awareness. For example, Lacey (1986) proposes that one of the functions of binge eating is to moderate or
‘block’ negative emotional states. He highlights a number of predisposing factors that might lead people to use eating in
377E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy
this way, including difficult family and peer relationships, a reliance on external standards for judging self-worth, and
family preoccupation with appearance, weight, and eating. Lacey (1986) suggests that these kinds of early experiences
typically lead to inadequate adult coping skills, including impaired interpersonal relationships and feelings of
ineffectiveness and lack of control. In this context, negative life events such as sexual conflict, changes in life
circumstances, or loss, result in excessive self-reflection and self-criticism. Because their attention is already focused
on food (as a result of both familial and cultural pressures for thinness), the manipulation of food becomes a convenient
defence mechanism. He proposes that in the light of external stressors, these women initially instigate dieting to try to
increase their self-worth by altering their body shape. However, as a result of this restriction they become more
vulnerable to binge eating, and consequently to vomiting. Despite finding these behaviours distasteful, bulimics may
continue to engage in them because they discover that they can be used both to fill an emotional void (e.g., associated
with boredom or loneliness) and to reduce emotional arousal (such as anxiety or distress). Indeed, women with bulimia
nervosa have reported that bingeing and vomiting can be used to “anaesthetize” negative emotional states
(Vanderlinden & Vandereycken, 1997). Given their lack of alternative coping skills, Lacey (1986) suggests that bulimic
women may therefore come to rely on bingeing and vomiting as ways of moderating their emotions.
Root and Fallon's (1989) model has many similarities to that proposed by Lacey (1986), though it primarily
considers bulimic individuals who are victims of abuse. Amongst other things, they suggest that bingeing is used as a
method of dealing with stress and tension, since these individuals lack sufficient alternative coping strategies. They also
suggest that bingeing can be used to temporarily anaesthetize powerful negative emotions and block out distressing
memories (particularly of abuse). Thus both blocking models see bingeing as allowing the individual temporary relief
from aversive self-awareness, in the context of deficient alternative coping strategies and feelings of lack of control.
3.2.3. Links with dissociation
Both cognitive narrowing and blocking are potentially related to the construct of dissociation. Dissociation is
described in DSM-IV as: “a disruption in the usually integrated functions of consciousness, memory, identity, or
perception of the environment” (American Psychiatric Association, 1994). For example, Schulman (1991) compared
the experience of bingeing to dissociation, and suggested that both function to block out painful emotions and
experiences. Binge eating and dissociation might therefore be alternative means of achieving the same end (i.e.,
emotional blocking). On the other hand, Demitrack, Putnam, Brewerton, Brandt, and Gold (1990) propose that
dissociative experiences lead to a narrowing of cognitive focus and the exclusion of important (often conflicting)
information from consciousness. Meyer, Waller, and Waters (1998) suggest that this style of thinking can lead to
behaviours that would usually be inhibited (e.g., eating) becoming disinhibited. As they point out, this has obvious
overlaps with the ‘escape from awareness’ model proposed by Heatherton and Baumeister (1991). Indeed, Heatherton
and Baumeister themselves make the connection between cognitive narrowing and dissociation. They cite research
from Abraham and Beumont (1982), which found that 75% of bulimics reported feelings of depersonalization and
derealization (i.e., aspects of dissociation) during binges. In addition, 72% of the women in this study reported
experiencing a reduction in negative mood states as a result of the binge. Taken together, these findings suggest that
dissociation might be an important construct to consider when investigating the processes of ‘blocking’ and ‘cognitive
narrowing’ described above.
In summary, binge eating may occur as a result of the disinhibition accompanying cognitive narrowing (a strategy
directly involved in the reduction of self-awareness by providing a blocking process. Both these strategies may have
connections with the construct of dissociation. Indeed, whilst these models describe different cognitive processes, they
are not incompatible with one another. In fact, McManus and Waller (1995) suggest that cognitive narrowing might be
particularly important in the initial onset of binge eating (along with dietary restriction), but once the blocking effects of
bingeingarediscovered,thisprocessmightbemore importantinexplaining itscontinued use.Ofcourse,there mayalso
be some individual differences in the relative importance of these processes in driving binge eating. Fig. 3 demonstrates
how these cognitive avoidance strategies might potentially connect eating disorders and anxiety disorders in thecontext
of specific cognitions.
In this model, disordered eating and anxiety behaviours are each seen as being associated with specific underlying
cognitions (i.e., related to eating, weight, and shape or to immediate threat cognitions). The use of cognitive narrowing
378E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386
Author's personal copy Download full-text
is seen as being driven by attempts to reduce the experience of anxiety, and disordered eating follows as a consequence
(i.e., as a result of reduced inhibitions). Conversely, blocking starts with disordered eating, since these behaviours are
directly used to block out the experience of anxiety. Fairburn et al. (2003) noted that the use of various blocking
behaviours (e.g., substance misuse, self-harm, or bingeing) is likely to be reinforced by the reduction in anxiety they
bring about. However, the underlying cognitions (i.e., related to eating, weight, and shape or to vulnerability to harm)
are not actually challenged. As such, it can be hypothesised that where these cognitive avoidance strategies are used,
the connection between anxiety and disordered eating will be maintained in the long term.
Of course, whilst the models for cognitive narrowing and blocking have been developed to account for bulimic
eating behaviours (i.e., bingeing and vomiting), it is also possible that similar processes apply to restriction. For
example, in a sample of women with anorexia nervosa (restricting subtype), Hinrichsen et al. (2003) found that higher
levels of social anxiety were associated with higher levels of dissociation (a construct which, as described above, has
been related to cognitive narrowing and blocking). Investigating how these cognitive strategies might apply to
restrictive eating behaviours might therefore be an important area for further research and theory development.
4. Aetiological determinants of the eating-anxiety link
Inherentinthese modelsistheprinciplethat having ananxiety disorder potentiallyincreasesthelikelihoodofhaving
a comorbid eating disorder (and vice versa). This is because both are hypothesised to be related to the same aetiological
factors (underlying harm avoidance cognitions). As indicated in relation to Fig. 1c, family study designs are one way to
and related methodologies). To test for shared familial risk factors, a four-group family design would ideally be
employed (eating disorder probands with anxiety disorders, eating disorder probands without anxiety disorders, non-
eating disorder probands with anxiety disorders, and non-eating disorder probands without anxiety disorders — i.e.,
both disorders would be expected among the relatives of the first three groups (compared to the non-clinical controls).
Unfortunately, to our knowledge, a study using these four groups (with balanced samples in each) has yet to be
completed. However, there are studies that have investigated the patterns of anxiety disorders amongst relatives of
eating disorder probands. Results from these studies currently yield mixed results. For example, Lilenfeld et al. (1998)
found that compared to controls, relatives of both anorexia and bulimia nervosa probands had significantly increased
rates of GAD and panic disorder. Furthermore, relatives of anorexia nervosa probands had significantly increased rates
of social phobia and OCD. However, although they found “some suggestion” of shared familial risk between social
phobia and anorexia nervosa, they report finding no evidence of shared causation between OCD and eating disorders
(i.e., the rate of OCD was only elevated amongst the relatives of eating disorder probands who themselves had OCD).
However, Lilenfeld et al. concede that their relatively small sample size (particularly in the anorexia nervosa group)
may have limited the power of their analyses.
Fig. 3. Cognitive avoidance strategies as a mechanism of association between anxiety and eating disorders.
379 E. Pallister, G. Waller / Clinical Psychology Review 28 (2008) 366–386