Behavioural and Cognitive Psychotherapy, 1997, 25, 113-145

Cognitive Theory in Anorexia Nervosa and Bulimia Nervosa: A Review Myra Cooper Warneford Hospital, Oxford This paper reviews cognitive theories of eating disorders as they are usually applied in treatment. More recent theoretical contributions and theory that is not widely applied are also reviewed. A set of hypotheses is derived from these theories and evidence for the validity of each hypothesis is discussed: this includes evidence from treatment studies, questionnaire studies and from experimental psychology. Following review of existing evidence, the paper summarizes the current status of cognitive theory in eating disorders. It then considers ways in which theory and research could be developed in order to improve and extend our understanding of cognitive content and processes in eating disorders. Limitations of existing models are highlighted and gaps in our knowledge, including knowledge of variables that typically have a central role in cognitive theory, are identified. Further strategies to test the validity of hypotheses derived from cognitive theory are suggested, together with strategies that might extend existing theory. It is concluded that much further research is needed, both to test the validity of existing theoretical contributions and to extend theory so that it will be more useful in clinical practice. Introduction Two eating disorders, anorexia nervosa and bulimia nervosa, are the subject of this paper. The former has a long history, with first descriptions by Sir William Gull and Ernest Lasegue in the nineteenth century. The latter has only come to the attention of psychiatry and the general public in the last sixteen years. Cognitive theories have been developed for both disorders: these theories will be discussed below. Cognitive theories of eating disorders Reports by clinicians of widely differing theoretical orientations have highlighted the existence and importance of a cognitive disturbance in anorexia nervosa (e.g. Bliss & Branch, 1960; Bruch, 1973; Selvini-Palazzoli, 1974). This disturbance is characterized by unusual beliefs about food, eating, Reprint Requests to Myra Cooper, Isis Education Centre, Warneford Hospital, Headington, Oxford OX3 7JX, UK. ® 1997 British Association for Behavioural and Cognitive Psychotherapies

114

M. Cooper

weight and shape. More recently, a similar disturbance has been identified in bulimia nervosa (e.g. Russell, 1979; Mitchell & Pyle, 1982). These clinical observations and anecdoctal reports have provided the basis for cognitive theories of eating disorders. In anorexia nervosa the cognitive theory developed by Garner and Bemis (1982) is the one that is most frequently used to guide treatment. In bulimia nervosa the theory developed by Fairburn and colleagues (Fairburn, Z. Cooper, & P.J. Cooper, 1986), which is based on that described by Garner and Bemis for anorexia nervosa, is most frequently used. The key elements of the two theories will be described below, followed by a description of the theoretical contributions made by two further papers (Guidano & Liotti, 1983; Vitousek & Hollon, 1990) that extend our understanding of the elements that may need to be included in a detailed cognitive model. Garner and Bemis Garner and Bemis base their theory on Beck's model of the emotional disorders, particularly depression (Beck, 1976; Beck, Rush, Shaw, & Emery, 1979). The theory takes as its fundamental premise the patients' belief "I must become thin". Garner and Bemis provide an account of the proximal sources of this belief and trace its influence on the patient's subsequent behaviour. The potential anorexia nervosa patient is typically introverted, sensitive and isolated and tries to live up to the expectations of others. She becomes withdrawn, feels helpless and perhaps depressed. Family and cultural influences give her the idea that losing weight will alleviate her distress. Initial successful weight loss results in feelings of self-control and is positively reinforced by others. Soon it is followed by a fear of weight gain and her behaviour becomes increasingly controlled by bizarre internal contingencies until neither the beliefs or the behaviour are affected, even by extremely punishing environmental consequences. Isolation from experience that might alter the beliefs increases vulnerability to the influence of self-generated distorted perceptions. Therapy follows lines suggested by Beck for the treatment of depression with some adaptations to help develop initial motivation for treatment. It includes tackling the belief that thinness is a value of inestimable worth. Garner and Bemis emphasize the discovery of distortions in the processing and interpretation of events, the evaluation of automatic thoughts or selfstatements and then of underlying assumptions, which can be inferred from self-statements. Some of these support avoidance of weight gain and eating while others support the desirability of weight loss and of dieting. Garner and Bemis give examples of distorted information processing and

j j I

Cognitive theory in eating disorders

115

typical assumptions. For example, patients tend to assume that events are black or white, right or wrong, good or bad. Typical assumptions include the belief that a perfect balance can and should be sought between disparate elements, that weight, shape or thinness can serve as the sole referent for inferring personal value and self-worth and that complete self control is desirable. Fairburn and colleagues

As in Garner and Bemis's account of anorexia nervosa, patients' attitudes toward their shape and weight are central to the maintenance of bulimia nervosa. Attitudes to food and eating are thought to be secondary consequences of attitudes to shape and weight. Like patients with anorexia nervosa, patients with bulimia nervosa tend to evaluate their self-worth in terms of their shape and weight. They view fatness negatively and thinness and self control positively. These attitudes are implicit and are based on unarticulated rules by which patients assign meaning and value to their experiences. They are dysfunctional because they are rigid and extreme and hold excessive personal significance. These typical beliefs and values reflect the operation of certain dysfunctional styles of reasoning or disturbances in information processing, similar to those described for anorexia nervosa. They include dichotomous thinking, overgeneralization and errors of attribution, for example, the belief that foods can be simply categorized as "fattening" or "nonfattening". Beliefs and values are reflected in thoughts that explain patients' behaviour such as frequent weighing or, alternatively, active avoidance of weighing. The reduction in food intake may also be directly attributed to thoughts concerning shape and weight. The theory describes a cognitive link between strict dieting and episodes of overeating. Intense concern with shape and weight leads patients to adopt extreme dieting rules that are impossible to obey. Inevitable minor deviations from these self-imposed rules are seen as catastrophic and evidence of weakness, reflecting a tendency to dichotomous thinking. As a result patients temporarily abandon all controls over eating and episodes of binge-eating occur. Guidano and Liotti Guidano and Liotto's theory is concerned primarily with anorexia nervosa (Guidano & Liotti, 1983). It is based on Bruch's observation that there are three fundamental deficits in the disorder. These are "a disturbance of delusional proportions in the body image and body concept"; "a disturbance in the accuracy of the perception or cognitive interpretation of stimuli

116

M. Cooper

arising in the body (particularly hunger and satiety)", and "a paralysing sense of ineffectiveness which pervades all thinking and activity" (Bruch, 1973, pp. 251-254). For Guidano and Liotti these deficits are all expressions of problems in cognitive structures relating to personal identity. Cognitive structures are beliefs and rules under which individuals operate and around which they organize their lives. Personal identity in anorexia nervosa is characterized by beliefs of general ineffectiveness and failure. Also typical is the belief that it is dangerous or useless to tell others about feelings or opinions and the expectation of rejection or criticism. In addition, patients also seem to have an imprecise attributional style. Unlike Garner and Bemis, Guidano and Liotti provide a detailed account of developmental factors. These involve a failure to develop autonomy, individuality and self-expression in childhood. This leads to beliefs about the impossibility of defining responsibility for one's own actions. Early disappointment in the child's emotional bond with her father leads to later reluctance to become involved in relationships for fear of similar rejection. The basis of the cognitive organization in anorexia nervosa is what Ellis would call a "dire need for love". Love is the saving idea, but as well as being a source of safety in the past it has also been a source of disappointment. To guard against disappointment those with anorexia nervosa demand that others be perfect before making any commitment. Equally they themselves must also be perfect to live up to the idealized other. Reality will, however, always prove disappointing. Because of an imprecise attributional style patients have difficulty identifying the real nature of the problem and this, combined with the particular personal identity structures, results in a maladaptive solution—dieting and weight loss. In therapy weight gain and the alteration of eating patterns is only a superficial goal. The deeper personal identity structures, the attitudes to reality, need to be corrected. The model emphasizes intervention at the level of underlying beliefs and rules and little attention is paid to issues of food, eating and weight or to the specific problems encountered in anorexia nervosa. This is consistent with their assumption that all eating problems share the same underlying cognitive structure. Vitousek and Hollon Vitousek and Hollon attempt to provide a framework to guide research into existing cognitive models rather than develop a new theory (vitousek & Hollon, 1990). They conceptualize eating disorders in terms of schema theory and, in the process, draw attention to low self worth, attitudes to

Cognitive theory in eating disorders

117

eating, weight and shape and evaluation of the self in terms of weight and shape. Specifically, these three aspects map onto disturbance at the level of self-schemata, weight-related schemata and weight-related self-schemata. Vitousek and Hollon propose that patients with eating disorders develop organized cognitive structures around issues of weight and its implications for the self that influence perception, thought, affect and behaviour. These weight-related self-schemata represent the core psychopathology of eating disorders and account for the persistence of both anorexia nervosa and bulimia nervosa, e.g. by determining selective attention and memory. In addition, they function to simplify, organize and stabilize the individual's experience of herself and the environment. This may be particularly important in patients with eating disorders whose experience of themselves and the environment may seem formless and chaotic. Conclusion The two cognitive models that are most frequently used to guide treatment in eating disorders (Garner & Bemis, 1982; Fairburn et al., 1986) assign unusual beliefs and attitudes a causal role in the maintenance of anorexia nervosa and bulimia nervosa. These beliefs and attitudes are concerned with the importance and meaning attached to weight and shape. The theories highlight the importance of three features: (1) self-statements or automatic thoughts; (2) dysfunctional styles of reasoning or disturbance in information processing; (3) underlying assumptions, beliefs or attitudes. Both are primarily models of maintenance. Guidano and Liotti (1983) and Vitousek and Hollon (1990) draw attention to three other elements that may be important in patients with eating disorders and which may need to be incorporated into a cognitive theory. These are: (1) personal identity structures or self-schemata, the content of which appears to refer to what Young (1990) has called core beliefs. In addition, Guidano and Liotti emphasize the importance of, (2) developmental processes and early experience. Vitousek and Hollon's paper is also important in drawing attention to the interdependence of structure and process (e.g. Landau & Goldfried, 1981) and, in particular, extends the importance of schematic processing in eating disorders beyond processes driven by assumptions to (3) processes driven by core beliefs. Early evidence for cognitive theories The four theoretical contributions described above were developed primarily on the basis of clinical observations and anecdotal reports and not, in general, on empirical evidence. These two sources of evidence can provide only limited support for cognitive theories of eating disorders. For example,

118

M. Cooper

there are no control groups, which makes it impossible to determine which aspects of the cognitive disturbance or developmental history are unique to patients with eating disorders and which may also be reported by young women without eating disorders, particularly those who are dieting or those who are depressed. Reports based on clinical interviews are inherently reactive, which means that relevant cognitions, processes or developmental factors may be missed. Reports are usually obtained retrospectively, and thus may be distorted accounts of what actually happened. They do not provide evidence to support or disconfirm the fundamental assumption made by cognitive theorists that cognitions have a causal role in the development or maintenance of the disturbed behaviour, both dietary restraint and episodes of bingeeating. More recently, empirical evidence has been collected to test different aspects of cognitive theories. A set of hypotheses or predictions can be derived from the four theoretical contributions discussed above: these will be presented below and, in the following section, they will each be considered in the light of relevant empirical evidence. Hypotheses derived from cognitive theories

At least nine predictions can be derived. These are as follows: (1) treatment based on the models, i.e. cognitive therapy, will be effective; (2) self-statements or automatic thoughts will reflect concern with food and eating, weight and shape; (3) underlying assumptions reflecting concern with food and eating, weight and shape will be strongly endorsed; (4) core beliefs will reflect global negative evaluation of the self; (5) dysfunctioal styles of reasoning or information processing errors and biases will be found in food and eating and in weight and shape concerns; (6) there will be a causal relationship between underlying assumptions and self-statements and eating behaviour, particularly dietary restraint; (7) dietary restraint, mediated by dichotomous thinking, will result in episodes of binge-eating; (8) schema driven processes will be evident in areas of core belief concerns; (9) early experience will be important in the formation of core beliefs. In the next section evidence that supports each of these hypotheses will be considered, taking each hypothesis in turn.

\

!i

| • j '

Cognitive theory in eating disorders 119 Evidence for hypotheses derived from cognitive theories Hypothesis 1: Treatment based on the models, i.e. cognitive therapy, will be effective Recent years have seen an increasing interest in cognitive behavioural treatments for eating disorders, particularly for bulimia nervosa. In so far as these are effective treatments, they provide one (indirect) source of evidence for the existence and importance of a cognitive disturbance in eating disorders. Bulimia nervosa. In bulimia nervosa initial studies were uncontrolled and single or multiple case reports (e.g. Fairburn, 1981). More recently, controlled studies have compared cognitive behaviour therapy with no treatment or waiting list groups (e.g. Freeman, Barry, Dunkeld-TurnbuU, & Henderson, 1988) and with other active treatments (e.g. Fairburn, Kirk, O'Connor, & P.J. Cooper, 1986). Some studies have also attempted to determine which of the different components of standard cognitive behavioural programmes are crucial to change (e.g. Fairburn et al., 1991). Controlled studies typically find significant decreases in bingeing and vomiting and improved scores on measures of general psychopathology in those who are treated with cognitive behaviour therapy. Many of these studies show that symptoms remain unchanged in those who are not treated (e.g. Agras, Schneider, Arnow, Raeburn, & Telch, 1989). A few studies have compared cognitive behaviour therapy with other active treatments. While some of these find a better outcome for those treated with cognitive behaviour therapy (e.g. Kirkley, Schneider, Agras, & Bachman, 1985) other studies have not confirmed this finding (e.g. Fairburn et al., 1991). Most of these studies have used behavioural techniques in both of the treatments being compared and the use of similar techniques might explain the rather similar outcomes. This suggests that cognitive disturbance may be rather unimportant. However, one study that included a measure of distorted attitudes to shape and weight (Fairburn et al., 1991) indicates that cognitive behaviour therapy may have a specific beneficial effect beyond that attributable to the use of behavioural techniques or indeed to techniques that make no reference to patients' eating disorder. This study provides some (indirect) support for the theoretical assumption that cognitive distortions are of primary importance in the maintenance of the disorder. Anorexia nervosa. There has been little interest in applying cognitive behavioural techniques to the treatment of anorexia nervosa. There are some uncontrolled case reports (e.g. P.J. Cooper & Fairburn, 1984) but only one controlled trial has been reported in the literature (Channon, de Silva,

i ^

'

120

M. Cooper

Hemsley, & Perkins, 1989). In this study no differences were found between cognitive behaviour therapy and behaviour therapy. However, the sample size was small (12 patients in each condition), and, given this and the general lack of studies, it seems most appropriate to conclude that the efficacy of cognitive therapy in this group of patients remains largely untested. Conclusion. Cognitive behaviour therapy is, within a stepped care approach, now widely regarded as the treatment of choice for more severe bulimia nervosa (Royal College of Psychiatrists, 1992). The success of cognitive therapy provides some support for cognitive theory. Although the evidence is mostly indirect, it includes support for the existence and importance of a cognitive disturbance. The evidence is based upon successful treatment of patients with bulimia nervosa. However, the evidence discussed above is limited. Different treatment studies have focused on different aspects of the cognitive disturbance. Few studies have assessed the disturbance directly and it is not yet clear which aspects are most important and which need to be changed for treatment to be successful. The evidence does not, in general, support specific elements of the theory. It would be helpful to clarify which aspects of content (at the level of self-statements or automatic thoughts and attitudes or beliefs) and information processing are disturbed in both eating disorders and which are most closely related to poor outcome and relapse. As well as informing theory, this might help to make treatment for the two disorders more effective. With respect to treatment, it is important to bear in mind that, despite significant decreases in bingeing and vomiting with cognitive behaviour therapy, many patients with bulimia nervosa remain with symptoms at the end of treatment and at follow-up. For example, a recent follow up study of patients who had received cognitive behaviour therapy (mean length of follow-up 5.8 years) found that, at follow-up, 37% met DSM-IV criteria for an eating disorder (American Psychiatric Association, 1994) and that only 50% were symptom free (Fairburn et al., 1995). There is currently very little evidence from treatment studies to support the validity of cognitive theory in anorexia nervosa. Hypothesis 2: Self statements or automatic thoughts will reflect concern with food and eating, weight and shape The content of self-statements has been assessed using self report questionnaires and techniques from experimental psychology. Only studies investigating specific psychopathology will be considered here. Questionnaire studies. Two studies have assessed self-statements or auto-

Cognitive theory in eating disorders 121

matic thoughts,1 one in bulimia nervosa (Phelan, 1987) and one in anorexia nervosa (D.A. Clark, Feldman & Channon, 1989).2 Phelan devised the Bulimic Thoughts Questionnaire, a three factor measure (self-schema, self-efficacy and salient beliefs) of frequency of thoughts about eating, weight and shape. On all three subscales patients with bulimia scored more highly than female controls. One subscale (salient beliefs) also distinguished patients with bulimia from obese women, with the patients obtaining a higher score. Clark and colleagues (D.A. Clark et al., 1989), using a modified version of the Distressing Thoughts Questionnaire (e.g. D.A. Clark & de Silva, 1985), found that patients with eating disorders (anorexia nervosa and bulimia nervosa) had a higher frequency of negative weight-related cognitions (including cognitions about food and eating, weight and shape) than female controls. In addition, these cognitions were more emotionally intense, uncontrollable, guilt-including and plausible in the two patient groups than in the control group. Experimental psychology. Self-statements have also been investigated in eating disorders using techniques from experimental cognitive psychology. One study has used "thought sampling" to examine self-statements (Zotter & Crowther, 1991), another has used concurrent verbalization or "thinking aloud" (M.J. Cooper & Fairburn, 1992a) while a series of studies conducted by one team of investigators (e.g. Leintenberg, Rosen, Gross, Nudelman & Vara, 1988) has used an abbreviated version of the thinking aloud technique. The studies conducted by Leitenberg and colleagues simply used concurrent verbalization as part of a battery of assessment measures and they did not aim to answer any research questions. In the study described by Zotter and Crowther (1991), subjects self monitored their thoughts every thirty minutes for two randomly selected 1. The concepts of automatic thoughts, underlying assumptions, core beliefs and schemas are, for the purposes of this paper, defined using criteria described elsewhere (e.g. Padesky & Greenberger, 1995; Beck, 1996). Automatic thoughts are thus situational specific, "momentto-moment, unplanned thoughts" (Padesky & Greenberger, 1995, p. 5); underlying assumptions are cross-situational beliefs or rules, including "should" statements and conditional "if . . . then" beliefs; core beliefs are absolute and dichotomous beliefs about the self, others and the world, while schemas refer to cognitive structures that guide processing, including both underlying assumptions and the structures in which core beliefs are embedded. In common with the usual use of terms within cognitive theory and therapy, the term information processing is used to refer to processing driven by assumptions while the more general term schematic processing is used for processing driven by structures containing core beliefs. 2. A third study describes the development of a "Bulimic Automatic Thoughts Test" (Franko & Zuroff, 1992). However, the rationale for selecting items, and examples given, suggest that the questionnaire is best considered primarily as an assessment of dysfunctional styles of reasoning.

122

M. Cooper

days. Subjects with bulimia nervosa reported a greater proportion of eatingand weight-related thoughts than female control subjects but this was no different from the proportion of eating- and weight-related thoughts in repetitive dieters. However, when the content of these cognitions was examined in more detail, subjects with bulimia nervosa had a greater proportion of weight- and shape-related cognitions than both dieters and female controls, both of whom had relatively more eating-related thoughts. M.J. Cooper and Fairburn (1992a) collected information on self-statements while subjects performed three behavioural tests. They included patients with eating disorders, two groups of dieters and normal controls. In addition to thinking aloud, subjects also completed a brief Thoughts Check-list, a measure of thoughts judged likely to occur in the three tasks to those with eating disorders. Each thought on the check-list was rated for frequency, duration and plausibility or belief. The concurrent verbalization results showed that patients with eating disorders had more negative selfstatements about eating, weight and shape during the tasks than the normal controls, while dieters occupied an intermediate position. The study also identified differences between patient groups. In patients with anorexia nervosa, concern with eating distinguished them most from dieters, while in bulimia nervosa concerns with weight and appearance distinguished the most from these two groups. Within group comparisons indicated that the pattern of concerns in the normal controls and dieters was similar to that found in the patients with bulimia nervosa and different from that found in the patients with anorexia nervosa. This suggests that there may be an extension of concerns found in the general population in bulimia nervosa while in anorexia nervosa concerns and preoccupations may be qualitatively different. Ratings obtained from the Thoughts Check-list, i.e. overall score (duration x frequency), absolute number of thoughts, frequency, duration and belief, did not discriminate groups as effectively as the thinking aloud data. Conclusion. These studies provide support for the existence of one of the key features of cognitive disturbance contained in the theories developed by Garner and Bemis and by Fairburn and colleagues. They suggest that there is a disturbance in cognition at the level of self-statements or automatic thoughts: automatic thoughts typically contain high levels of thoughts about food and eating, weight and shape. Unlike clinical and anecdotal reports, these studies all include control groups, demonstrating that different groups of subjects differ in response in predicted directions. This provides information on concurrent (criterion related) validity. However, in all studies (those using self-report questionnaires and those using techniques from experimental cognitive psychology) information on

Cognitive theory in eating disorders 123

reliability and validity is limited. In addition to criterion related validity, Phelan (1987) provides some preliminary data on the factor structure of her questionnaire but no data on the reliability or validity of the three factors identified. Clark and colleagues (D.A. Clark et al., 1989) provide preliminary data on the internal consistency and test retest reliability of their measure, both of which reach acceptable levels, but no other information. Studies using less structured techniques (Zotter & Crowther, 1991; MJ. Cooper & Fairburn, 1992a), provide information on inter-rater reliability for coding subjects' responses. For all categories in both studies this reaches acceptable levels. No other information is provided. The questionnaire measures, in particular, have some further limitations. Only one appears to have been developed on the basis of empirical research (Phelan, 1987). In this study thoughts diaries kept by patients with bulimia nervosa were used to derive questionnaire items. In addition, while both questionnaire studies compared patients with normal, non eating disordered females, only one (Phelan, 1987) included an additional relevant control group, obese women: other appropriate comparison groups such as restrained eaters, those who are dieting to lose weight, those who show symptoms of an eating disorder but who do not meet full diagnostic criteria and those who are depressed have not been studied. More generally, although it is not a necessary feature of self-report questionnaires, both of the questionnaire measures are highly structured and thus may miss features of interest which the investigator has not thought to ask about, particularly if they lack an empirical base. They may, therefore, simply confirm the investigator's existing hypotheses without providing a full picture of the cognitive disturbance. They also ask for retrospective reports and their accuracy in capturing situation specific, moment-tomoment thoughts is, therefore, unclear. In contrast, thought sampling and concurrent verbalization are relatively unstructured techniques. There are no probe questions and both yield immediate information on thoughts that occur in a particular situation. They are, therefore, more likely to provide an accurate record and less likely to miss features of interest that the investigator has not yet considered. Existing studies using relatively unstructured techniques have also included dieters, an important control group, as well as female controls although, like the questionnaire studies, they have not included a depressed control group. This would appear to be a particularly important group if, as some studies suggest (e.g. M.J. Cooper & Fairburn, 1992a), the valence of thoughts relevant to concern with food and eating, weight and shape, best distinguishes eating disordered patients from normal controls. In addition, it might also be important to assess the presence of negative self-statements unrelated to patients' specific concerns.

124

M. Cooper

Thought listing and concurrent verbalization techniques do, of course, have their own limitations: they are time consuming for those who take part and for researchers. Subjects may also find thought sampling in daily life to be unacceptably intrusive. Nevertheless, both are particularly useful techniques in a relatively new area of research: they may provide a sound empirical basis for the development of measures, including questionnaire measures, that are rather easier to administer. The very detailed information gathered using concurrent verbalization, on all thoughts in key situations likely to be of concern to patients with eating disorders may, as in one of the studies described here (M.J. Cooper & Fairburn, 1992a), be particularly useful in helping researchers to begin to identify differences in cognition between patients with anorexia nervosa and patients with bulimia nervosa. Existing, more structured measures that lack an empirical basis may not be sensitive to such differences. Hypothesis 3: Underlying assumptions reflecting concern with food and eating, weight and shape will be strongly endorsed Endorsement of underlying assumptions hypothesized to be uniquely characteristic of eating disorders appears to have been assessed empirically only by self report questionnaire. Self-report questionnaires. Four measures have been developed (ReynaMcGlone, Ollendick, & Hart, 1986; Scanlon, Ollendick, & Bayer, 1986; Schulman, Kinder, Powers, Prange, & Gleghorn, 1986; Mizes, 1988)3 All have been used to assess cognition in bulimia nervosa. In addition, the measures developed by ReynaMcGlone and colleagues (ReynaMcGlone et al., 1986) and by Mizes (Mizes, 1992) have also been used to assess cognition in anorexia nervosa. Studies have found that responses on all four questionnaires distinguish those with eating disorders from non eating disordered control subjects, with the eating disordered scoring more highly than the controls on agreement with a variety of assumptions concerned with food and eating, weight and shape on three measures (Attitude and Belief Survey: Scanlon et al., 1986; Bulimia Cognitive Distortions Scale: Schulman et al., 1986; Powers, Schulman, Gleghorn, & Prange, 1987; Anorectic Cognitions Scale: Mizes, 1992) and scoring lower than the controls (reverse scoring system, rating belief in self-efficacy) on one measure (Self-Assessment Scale: Reyna McGlone et al., 1986). One study has also used an additional, psychiatric control group (Mizes, 1992). This study found that bulimia nervosa and 3. A distinction is made here between studies whose primary aim is to assess content and studies whose primary aim appears to be to assess the different types of dysfunctional styles of reasoning contained in the content.

Cognitive theory in eating disorders 125

anorexia nervosa patients scored more highly on agreement with assumptions than a mixed group of non eating disordered psychiatric patients. While ReynaMcGlone and colleagues found that restricting anorexia nervosa patients, unlike bulimia nervosa patients, scored within the normal range on the Self-Assessment Scale, Mizes (1992) found that patients with anorexia nervosa did not differ from patients with bulimia nervosa in scores on the Anorectic Cognitions Scale. This difference may be due to the rather different content areas assessed by the two measures. While Mizes' measure assesses assumptions relevant to all three areas of patients' concerns with food and eating, weight and shape, without providing separate subscales for each concern, the measure developed by ReynaMcGlone and colleagues assesses cognitions about eating, specifically ability to control eating. It may be that concern with control of eating is of particular importance to those with bulimia nervosa perhaps because, compared to those with anorexia nervosa, they experience relatively more episodes of loss of control over eating. Alternatively, it may be that different diagnostic criteria, including only restricting anorexics in one study, may explain the difference. Conclusion. These studies provide support for the existence of a second key feature of the cognitive disturbance contained in the theories developed by Garner and Bemis and by Fairburn and colleagues. They suggest that there is a disturbance in cognition at the level of underlying assumptions: assumptions reflecting a high degree of concern with food and eating, weight and shape are strongly endorsed by patients with eating disorders. Once again, unlike clinical and anecdotal reports, these studies all include control groups, demonstrating that different groups of subjects differ in their responses in predicted directions. This provides information on concurrent (criterion related) validity. No further information on reliability and validity is available for the Attitudes and Beliefs Survey. However, preliminary information is available for two measures and more detailed information for the third measure. The Self-Assessment Scale appears to have acceptable internal consistency and some preliminary data provide evidence of construct (convergent) validity. Preliminary findings with the Bulimic Cognitive Distortions Scale suggest two factors (cognitions associated with automatic eating behaviour and cognitions associated with physical appearance); acceptable internal consistency; evidence for construct (convergent and discriminant) validity and further evidence for concurrent (criterion-related) validity, using discriminant function analysis. The Anorectic Cognitions Scale has been researched in most detail (Mizes & Klesges, 1989; Mizes, 1990, 1991, 1992). Factor analysis suggests a replicable three factor solution (perception of weight and eating as a basis for approval from others, belief that rigid weight and eating control is fundamental to self worth and rigidity

126

M. Cooper

of weight and eating regulation efforts). Internal consistency of the items on each of the three factors and test-retest reliability is high. Construct (convergent and discriminant) validity has been demonstrated together with additional concurrent (criterion-related) validity in a normal population. Although normal, non eating disordered control groups have been employed in studies investigating all four measures, no study has assessed cognition using any of these measures in restrained eaters, dieters or those with some symptoms but not a diagnosis of an eating disorder. Like the questionnaire studies of self-statements they have not included a depressed control group or assessed belief in cognitions characteristic of those who are depressed and which contain content unrelated to food and eating or to weight and shape. No information is provided on how items were derived and selected for the Attitudes and Beliefs Survey, while items for the remaining three questionnaires were developed on the basis of knowledge of the literature and clinical experience: thus these measures do not have a basis in empirical research. Once again, the measures are highly structured and may, particularly without an adequate empirical base, be reactive, simply confirming hypotheses and missing cognitions that the investigator has not thought to include. No study has separated content areas clearly, particularly weight and shape from food and eating, despite suggestions elsewhere that these may be separate areas of concern to patients and, of theoretical importance, that food and eating concern may be secondary to weight and shape concern (Fairburn et al., 1986). Hypothesis 4: Core beliefs will reflect global negative evaluation of the self Only one, very preliminary, study appears to have attempted to assess core beliefs in patients with eating disorders (MJ. Cooper, Todd, & CohenTovee, 1996). Using a sentence completion task (Padesky & Mooney, 1993), this study found that patients with anorexia nervosa and bulimia nervosa recorded more negative self beliefs than normal controls. Inter-rater reliability appears to be acceptable but no other information on reliability or validity has been reported. The study also suggested that core beliefs about the world might be important: patients with eating disorders had fewer positive beliefs about the world than the normal controls. The method is quick and easy to administer and thus may be a particularly useful measure of core beliefs in clinical practice.

\ j

Cognitive theory in eating disorders 127 Hypothesis 5: Dysfunctional styles of reasoning or information processing errors and biases will be found in food and eating and in weight and shape concerns Styles of reasoning and information processing errors and biases, in areas of specific psychopathology, have been assessed using self report questionnaires and techniques from experimental psychology. Self-report questionnaires. Two questionnaires have been designed to assess dysfunctional styles of reasoning in patients' uniquely characteristic areas of concern with food and eating, weight and shape (Thompson, Berg, & Shatford, 1987; Franko & Zuroff, 1992).4 Both questionnaires included items that were judged to contain characteristic distortions (e.g. over-generalization, perfectionism, dichotomous thinking) and examples of items suggest that distortions were being assessed at the level of underlying assumptions.5 In both studies higher levels of distortions were found in patients with bulimia nervosa than in normal controls. One study assessed frequency of experiencing distortions (Franko & Zuroff, 1992) while it is not clear in the second study (Thompson et al., 1987) which dimension of distorted cognition (e.g. frequency, belief) was being assessed. One study found that distortions also discriminated those with "bulimic-like" symptoms from those with a diagnosis (Thompson et al., 1987). The other study found that those with bulimia nervosa also reported a higher frequency of distortions than depressed college students and obese subjects (Franko & Zuroff, 1992). Experimental psychology. One study, already discussed above, assessed cognitive distortions in addition to cognitive content when collecting data using thought sampling (Zotter & Crowther, 1991). Other studies have used paradigms from experimental psychology to assess selective attention and, to a much lesser extent, selective memory. Thought sampling data indicated that patients with bulimia nervosa were more likely to report dichotomous thinking relevant to their concerns with eating and weight than repetitive dieters and normal controls (Zotter & Crowther, 1991). This study appears to assess distortion at the level of selfstatements or automatic thoughts. Selective attention in eating disorders has generally been assessed using modified versions of the Stroop (1935) colour-naming paradigm (e.g. Channon, Hemsley, & de Silva, 1988; M.J. Cooper, Anastasiades, & Fair4. A third questionnaire was found but this was developed and has been used only with a sample of bulimics who did not meet diagnostic criteria (Dritschel, Williams, & P.J. Cooper, 1991). 5. Cognitive distortions may be evident in self-statements or automatic thoughts as well as in underlying assumptions.

128

M. Cooper

burn, 1992) while one study has used a dichotic listening task (Schotte, McNally, & Turner, 1990). Studies using a variety of modifications of the original Stroop task have shown that patients with anorexia nervosa are slower to colour-name words relevant to their concerns than normal controls (Channon et al., 1988; Ben-Tovim, Walker, Fok, & Yap, 1989; BenTovim & Walker, 1991; M.J. Cooper & Fairburn, 1992b; Green, McKenna, & de Silva, 1994; Long, Hinton, & Gillespie, 1994) and that patients with bulimia nervosa are also slower than normal controls (Ben-Tovim et al., 1989; Ben-Tovim & Walker, 1991; Fairburn, P.J. Cooper, M.J. Cooper, McKenna, & Anastasiades, 1991; M.J. Cooper & Fairburn, 1992b; M.J. Cooper et al., 1992). A few studies have used additional control groups. One study included two groups of dieters (M.J. Cooper & Fairburn, 1992b) and showed that disturbance in information processing occurred only in patient groups and in dieters who had some symptoms but not a diagnosis of eating disorder. Normal dieters were no different from the normal, non dieting controls. Patients with anorexia nervosa and patients with bulimia nervosa appear to differ from adolescent females who are concerned about their weight and shape (Ben-Tovim & Walker, 1991) but patients with anorexia nervosa did not, in another study, differ from obese restrained eaters (Long et al, 1994). Some studies have included words relevant to all three areas of patients' concerns with eating, weight and shape in a single colour-naming task (e.g. M.J. Cooper et al., 1992) while other studies have separated food and eating concerns from weight and shape concerns (e.g. Channon et al., 1988). The pattern of results obtained in some of the studies which have separated concerns suggests possible differences between patient groups in information processing. It seems that patients with anorexia nervosa and bulimia nervosa may be slower to colour name words related to food and eating while only those with bulimia nervosa may also be slower on shape related words (Channon et al., 1988; Ben-Tovim et al., 1989). This finding is compatible with the differences observed in the concurrent verbalization study (M.J. Cooper & Fairburn, 1992a). However, some more recent studies have produced rather different findings. One study found that patients with anorexia nervosa were slower to colour-name both types of word (Green et al., 1994) while another found that only patients with anorexia nervosa, and not those with bulimia nervosa, were slower to colour-name words related to concern with food and eating (Perpifla, Hemsley, Treasure, & de Silva, 1993). The study which used a dichotic listening task to measure selective attention found that patients with bulimia nervosa, compared to normal controls,

Cognitive theory in eating disorders 129

attended to the word "fat" more often than to a control word (Schotte et al., 1990). Two studies have assessed selective memory in patients with eating disorders (King, Polivy, & Herman, 1991; Sebastian, Williamson, & Blouin, 1996). The first was designed primarily to assess person memory in restrained eaters but also included a small number of eating disordered patients. The results suggested that the patients had better recall for weight and food related information than for other information (King et al, 1991). The other study found that a mixed group of eating disorder patients recalled more "fatness" related words than normal and weight preoccupied controls (Sebastian et al., 1996). Conclusion. The two self-report questionnaire discussed above have both assessed a wide range of distortions. Relevant control groups, those with a partial syndrome, obese subjects and depressed college students, have been included although a depressed patient group, restrained eaters and dieters without symptoms have not been studied. Thoughts diaries were used as one source of data for items in one study (Franko & Zuroff, 1992) while the other study appears to have relied on expert choice to select items (Thompson et al., 1987). Like most highly structured self-report measures they may be reactive and, when assessing distortions at the level of situation specific automatic thoughts, their accuracy may be unclear. Some additional data on the reliability and validity of the two self-report questionnaire measures are available. Details of internal consistency are provided for the Food and Weight Cognitive Distortions Survey (Thompson et al., 1987) and this reaches acceptable levels. More information is available for the Bulimic Automatic Thoughts Test (Franko & Zuroff, 1992), with generally acceptable split half reliability, good internal consistency, and evidence of construct (convergent and divergent) validity. The Stroop colour-naming task provides an objective measure of selective attention, one aspect of disturbed information processing. Like the concurrent verbalization technique it avoids the problems of reactivity and accuracy which may be associated with self-report measures. As noted above, several studies have demonstrated disturbed information processing using this technique, both in anorexia nervosa and in bulimia nervosa. Control groups of dieters (M.J. Cooper & Fairburn, 1992b) and obese retrained eaters (Long et al., 1994) have been used in some studies. However, there are methodological problems and problems with data analysis in many studies. In some studies different groups of stimuli have not been presented in a balanced order; thus order effects may explain the findings. Others also use unsuitable methods, usually separate one way analyses of variance, to analyse the data obtained from different groups of stimuli when, in fact,

130

M. Cooper

two way analysis of variance would be more appropriate. Although it seems that there may be differences in attentional bias between patients with anorexia nervosa and patients with bulimia nervosa the results are conflicting and possible differences need to be investigated in a single study design. Unlike existing designs this should include both groups of patients and both types of stimulus word (food and eating and body shape) and use the improved methodology employed more recently. In addition, it might be useful to separate out weight related words since these may be a separate concern for people with eating disorders (Fairburn et al., 1986). It is also important to note that, in normal subjects, disturbed processing of food related information may be closely related to hunger (Channon & Hayward, 1990). Thus, inclusion of food and eating related information and differences between the patients and controls in hunger, rather than eating disorder specific psychopathology, may explain findings with these stimuli (Vitousek & Hollon, 1990). However, it should also be noted that, in a clinical sample, no relationship was found between interference and selfreported hunger (Channon et al., 1988). There is some, limited, additional information on the reliability and validity of the Stroop paradigm in eating disorders beyond ability to distinguish different subject groups. Evidence relating to concurrent validity indicates that, using multiple regression analysis, frequency of purging, and not level of general psychiatric symptomatology, is the best predictor of interference (M.J. Cooper & Fairburn, 1993). However, one study showed that score on the Eating Attitudes Test (EAT: Garner & Garfinkel, 1979) did not correlate with colour-naming interference in patients with anorexia nervosa (Channon et al., 1988) and, while another (M.J. Cooper et al., 1992), showed that EAT score and interference were correlated in patients with bulimia nervosa, the relationship disappeared when level of depressive symptoms was taken into account. There appears to be only one study which has used a different attentional paradigm (Schotte et al., 1990). Further work is needed on such measures; only a normal control group has been used and there is no information about the reliability and validity of the measure beyond its ability to discriminate patients from normal subjects. Further work is also needed with memory paradigms in eating disorders. Control groups, restrained eaters and weight preoccupied subjects, have been included but no study has also included depressed subjects. The number of patients was small (six) in one study (King et al., 1991) and most appeared to have a diagnosis of anorexia nervosa. The method used to analyse the patient data in this study (visual inspection of results) makes the results difficult to interpret. Inclusion of food related information and

]

j \ j j j

Cognitive theory in eating disorders

131

differences between the patients and female controls in hunger, rather than eating disorder specific psychopathology, may also provide a more parsimonious explanation for the findings, although this remains to be demonstrated. In the second study (Sebastian et al., 1996) a depressed control group and generally negative words need to be added to the study design. All the fat related words included in this study appear to be negatively toned: at present it is unclear whether the bias found is, therefore, part of a more general negative bias due to co-existing depressive symptoms or an example of a more specific bias, one that is confined to eating disordered patients. Neither study investigated self referent recall and the relative contribution of positive and negative stimuli to selective recall, although both may be important (Vitousek & Hollon, 1990). Hypothesis 6: There will be a causal relationship between underlying assumptions and self-statements and eating behaviour, particularly dietary restraint Treatment follow-up studies and one experimental study provide some support for this hypothesis. Follow-up studies. The suggestion that cognitive disturbance has a causal role is supported by the finding, following treatment for bulimia nervosa, that residual level of attitudinal disturbance at the end of treatment predicts outcome, defined in behavioural terms, at 12 months (Fairburn, Peveler, Jones, Hope, & Doll, 1993). These are preliminary findings and a recent study failed to replicate them (P.J. Cooper & Steere, 1995). However, as the authors of this study note, this may well be because of restricted assessment of cognitive disturbance. They suggest that the Eating Disorder Examination (EDE: Z. Cooper & Fairburn, 1987), a semi-structured interview used to assess attitudes to shape and weight in both these studies, may not capture the cognitive variables that are changed by cognitive therapy. Additional support for the causal role of cognition comes from other bulimia nervosa treatment follow-up studies. For example, several studies find that gains made with cognitive behaviour therapy are more likely to be maintained than gains made with other treatments (e.g. Thackwray, Smith, Bodfish, & Myers, 1993). Most of these studies have relatively brief follow-up periods and support for the causal link is indirect: none of these studies has assessed cognition directly and none has investigated patients with anorexia nervosa. Experimental psychology. One study has been conducted, using experimental psychology techniques, which demonstrates a causal link between attitudes to eating, weight and shape and eating behaviour (M.J. Cooper, D.M Clark, & Fairburn, 1993). This study activated assumptions about eating, weight and shape in patients with bulimia nervosa. There was an

132

M. Cooper

increase in negative self-statements, following the experimental manipulation, in the experimental group but not in the control group. As predicted, it was found that patients whose assumptions were activated ate less in a taste test than patients whose assumptions were not activated. Thus, assumptions about eating, weight and shape appear to be causally related to disturbed eating, specifically to reduction in intake, in patients with bulimia nervosa. Conclusion: Little research has been conducted to investigate the causal role of assumptions relevant to patients' concerns. Preliminary reports generally support the hypothesis: failure to replicate supporting findings in one study may be due to limited assessment of relevant cognitions. Hypothesis 7: Dietary restraint, mediated by dichotomous thinking, will result in episodes of binge-eating The relationship between dietary restraint and binge-eating was also investigated in the study described above (M.J. Cooper et al., 1993). Self-reported binges were calculated for the 24 hours following the experiment using a scheme based on that adopted in the EDE for distinguishing between objective and subjective bulimic episodes. Contrary to expectations, decreased food intake did not appear to lead to a greater number of objective bulimic episodes. However, those in whom assumptions had been activated reported fewer subjective bulimic episodes. The mediating role of dichotomous thinking does not appear to have been studied empirically. Conclusion. Several possible explanations exist for the findings in the study conducted by Cooper and colleagues (M.J. Cooper et al., 1993). These include the possibility that the reduction in food intake was not sufficient to lead to binge-eating; that the time course of the experiment was not long enough to demonstrate increased binge-eating and, not incompatible with either of these explanations, the possibility that because the experimental group ate more in the taste test they experienced a greater sense of loss of control over eating, a feeling that persisted for some hours. Further research is needed to investigate this hypothesis; in particular to investigate the role of cognition in mediating episodes of binge-eating. Hypothesis 8: Schema driven processes will be evident in areas of core belief concerns Young (1990) identifies three processes that are associated with schemata and that function to keep them unchallenged and unchanged. These are schema maintenance, schema avoidance and schema compensation processes. The hypothesis that schema driven processes such as these will be

Cognitive theory in eating disorders 133

evident in areas of core belief concerns does not appear to have been investigated empirically in eating disorders. Hypothesis 9: Early experience will be important in the formation of core beliefs Like Hypothesis 8, this hypothesis does not appear to have been investigated empirically in eating disorders. Current status of cognitive theories

The four theoretical contributions described above (Garner & Bemis, 1982; Fairbum et al., 1986; Vitousek & Hollon, 1990; Guidano & Liotti, 1983) identify many of the key elements of cognitive theory that have been described elsewhere in the literature on cognitive theory and therapy and apply them to eating disorders. While all contributions in eating disorders were initially based on clinical observation, empirical support for some of the elements has now been gathered. This is encouraging and generally supports various aspects of the theory. However, in general, the development of cognitive theory in eating disorders has lagged behind the development of cognitive theory in other disorders. Disappointingly little research has been conducted with the aim of testing hypotheses derived from cognitive models of eating disorders. In addition, while promising, much of the support discussed here remains preliminary or indirect and contains various problems. Most support exists for the components highlighted in the two theories that are commonly applied in treatment: some of the hypotheses contained in the additional contributions made by Vitousek and Hollon and by Guidano and Liotti do not appear to have been investigated at all. Further work is needed to test the predictions made by Garner and Bemis and by Fairburn and colleagues: at the same time this work clearly needs to be extended to test the additional predictions contained in the contributions made by Vitousek and Hollon and by Guidano and Liotti. But, as well as simply accumulating more evidence for existing theoretical predictions, some of which already have some empirical support, future research needs to be designed to extend all four theoretical contributions: all need further development in a variety of ways. Much more work needs to be done to develop existing theory further, and more detail provided, particularly if it is to be more useful to clinicians. The section below will discuss some possible future developments of cognitive theory in eating disorders together with areas for further research.

134

M. Cooper

Future development of cognitive theory and research

This will be considered under various headings. Treatment studies As has been noted above, treatment and treatment follow-up studies, which provide indirect evidence for the validity of cognitive theory, would benefit from improved assessment of cognitive variables: this would provide (indirect) clues about which aspects of cognition might be important in the maintenance of eating disorders. Aspects that are relatively more important might be expected to change most, and remain changed at follow-up, in those whose behaviour improves most with cognitive therapy. However, improved assessment of cognitive variables requires knowledge of which aspects of cognition are likely to be disturbed and adequate measures, i.e. providing a complete picture and psychometrically sound, for their assessment. This will be considered below. Measures of cognition Key elements identified empirically as important and disturbed in eating disorders include characteristic automatic thoughts, underlying assumptions and information processing. However, as has been noted elsewhere (Mizes & Christiano, 1995), little information on reliability and validity is available for many of the measures that have been developed to test the importance of these variables: this includes both self-report questionnaire measures and various experimental psychology techniques such as thinking aloud and thought listing. Further work is needed on most measures if they are to confirm the importance of the elements they assess; be of further use in research, including therapy outcome research; and be useful tools for conceptualizing individual problems and assessing individual therapeutic outcome. This is an important area for development. However, besides just taking existing measures and establishing their psychometric properties, researchers may wish to develop new measures, building on what has been learned about cognition in eating disorders more recently and taking criticisms of existing measures into account. For example, the one measure that has been extensively researched, the Anorectic Cognitions Scale (e.g. Mizes, 1988), does not distinguish different content areas of concern, despite suggestions that it may be important to do so theoretically (Vitousek & Hollon, 1990). With respect to automatic thoughts, it might be useful, as two studies have attempted (M.J. Cooper & Fairburn, 1992a; D.A. Clark et al., 1989), to investigate dimensions of thoughts other than simple frequency. In treatment, ratings of belief are often obtained and characteristic distortions are often identified and chal-

Cognitive theory in eating disorders 135

lenged. Only one study has investigated the presence of distortions in situation specific thoughts. Such measures could be used experimentally to answer research questions and thus develop theory as well as provide useful, psychometrically sound measures for use in other studies and in clinical practice. Underlying assumptions and core beliefs Beliefs and attitudes have been relatively neglected in eating disorders (Vitousek & Hollon, 1990), yet in many ways they form the heart of any cognitive model (Persons, 1989). They drive the self-statements that lie behind the disturbed behaviour and thus play an important part in a cognitive theory. However, despite their theoretical importance, few studies have focused on underlying assumptions or core beliefs and few studies have investigated the processes associated with them, for example selective memory (associated with underlying assumptions) or the schema driven processes described by Young (associated with core beliefs). The exception is the large literature on selective attention, using modified versions of the Stroop paradigm. Vitousek and Hollon, in particular, highlight the schematic properties of beliefs and attitudes. Like Guidano and Liotti they emphasize underlying attitudes and rules, particularly self-schemata (core beliefs) and weightrelated self-schemata (underlying assumptions). However, they do not elaborate on the nature of the schemata, on how they might be constructed or on how they might operate. How might underlying assumptions and core beliefs be investigated further? Self-report questionnaires and techniques from experimental cognitive psychology have already proved useful in assessing self-statements and information processing. There are several ways in which assumptions and beliefs could be investigated using similar techniques. It might be useful to investigate both the structure and content of beliefs as well as the different processes associated with them including, for example, aspects of information processing associated with underlying assumptions other than selective attention. Structure and content of assumptions and beliefs Techniques used to investigate semantic network theory (e.g. Collins & Loftus, 1975; Anderson, 1983) and network theory of affect (e.g. Gilligan & Bower, 1984) could provide some clues about the construction and content of cognitive structures that have schematic properties. For example,

136

M. Cooper

although, superficially, weight-related schemata,6 i.e. what it means to be fat and what it means to be thin, might appear to exist in a similar form in many people without eating disorders, in fact the relationship between the concepts fat and thin and other concepts might be rather different in patients with eating disorders. One possibility is that concepts of fat and thin are relatively well elaborated in patients with eating disorders and that there may be a greater number of links to other concepts. Number and type of links to other concepts could be measured using attribute listing such as that described by Landau (1980) for investigating the concept of dog in dog phobics. An adaptation of the contextual priming task (Fischler & Bloom, 1979) might also be used to investigate the relationship between different concepts. In addition, the degree of emotion associated with each concept or with individual relationships between concepts may be greater in patients than in normal controls, dieters or restrained eaters. This may be related to the accessibility of schemata, turning a belief into what Beck has described as a hot cognition. The content and structure and associated emotional links of weight-related self-schemata, as well as that of self-schemata or core beliefs, could also be investigated using these techniques. Little has been written about the exact content of assumptions and beliefs: the few examples of weight-related self-schemata given by Vitousek and Hollon appear to be at a superficial level, more typical of automatic thoughts than underlying meanings or beliefs. The a priori division of underlying assumptions into two types (Vitousek & Hollon, 1990) and empirically based division into three types (Mizes, 1988) also highlights the need to explore further, perhaps using self-report questionnaires and factor analysis, the precise content of underlying assumptions and, in particular, how they group into themes. An important variable to enter into such analyses may be the specific content area of patients' concern, i.e. weight and shape, food and eating, a variable whose importance is discussed further below. Information processing. Both general and specific aspects of information processing could be investigated using techniques from experimental psychology. At a general level, selective memory has not been investigated extensively in eating disorders Just as patients with eating disorders selectively attend to information relevant to their concerns, so might they selectively recall such information. Paradigms that have proved useful in assessing selective 6. Vitousek and Hollon (1990) identify weight-related schemata and weight-related selfschemata, both of which appear to be examples of underlying assumptions. As noted above, weight-related self-schemata are believed to represent the core psychopathology of eating disorders. However, as will be noted later, it is not clear precisely how the two structures relate to each other.

Cognitive theory in eating disorders 137

memory in depression (e.g. Derry & Kuiper, 1981) and anxiety disorders (e.g. Mogg, Mathews, & Weinman, 1987) could be used. Interpretative bias, the extent to which patients with eating disorders interpret different types of ambiguous events in terms of their concerns, is another information processing bias that might be found in eating disorders. It could be investigated using a questionnaire similar to that developed to measure the interpretation of ambiguous events in a negative fashion (Butler & Mathews, 1983). Such a bias could also be investigated using some of the more objective techniques employed in studies of anxiety (e.g. Mathews, Richards, & Eysenck, 1989). Techniques such as those used by Markus for investigating processes associated with weight-related self-schemata in non-clinical populations (Markus, Hamill, & Semis, 1987) could also be used. At a more fine grained level, specific biases could be investigated. For example, the questionnaires developed by Butler and Mathews could be adapted to investigate some of the specific errors in information processing observed clinically in patients with eating disorders, including errors of personalization and magnification (Garner & Bemis, 1982). Processes associated with self-schemata or core beliefs also need further investigation. Because processes function to maintain beliefs, techniques from experimental psychology might usefully be employed here too, including selective attention and selective memory paradigms. However, care would need to be taken in determining specific predictions because, for example, both processes of maintenance and avoidance may occur, driven by the same core belief, at different times. Schema driven processes (most notably and importantly, schema compensation) may also be reflected in underlying assumptions and their assessment may, therefore, overlap with that of underlying assumptions; thus they may also usefully be explored using self-report questionnaires, and factor analysis might be used to identify themes in processing. Conclusion. Appropriate control groups are necessary when using these techniques, including dieters, restrained eaters and those with some symptoms but not a diagnosis of an eating disorder. It may also be useful to investigate patients with depression. For example, self-schemata in patients with eating disorders (both the content or structure, i.e. core beliefs, as well as associated processes) may well be similar to that found in patients with depression. Detailed empirical investigation has not been carried out with assumptions and beliefs: it is possible that such work might reveal differences between patient groups. M.J. Cooper (1991) has suggested that differences identified between patients with anorexia nervosa and patients with bulimia nervosa in self-statements might be due to differences in

138

M. Cooper

content of beliefs and attitudes. Recently clinicians have emphasized the similarity of the cognitive disturbance in the two disorders. However it is also important to identify differences, not simply to explain why some may develop "restricting" anorexia nervosa while others develop a disorder with bulimic episodes, but also as a basis for developing more effective treatments for each disorder. To date, no theory addresses the question of why some patients with over-valued ideas concerning their shape and weight successfully maintain control over their eating and develop "restricting" anorexia nervosa. Possible differences between patient groups in assumptions and beliefs require further investigation Content specificity Of particular importance, and relatively ignored by researchers, is content specificity in cognition. Theory indicates that concern with food and eating may be secondary to concern with weight and shape (Fairburn et al., 1986). However, differences in content of cognition have not been investigated thoroughly in empirical studies: most studies combine different categories of concern into a single category. If concern with weight and shape is a feature of the core psychopathology, as suggested, for example by Fairburn and colleagues and by Vitousek and Hollon, then it may be very important to separate these concerns from food and eating concerns and to develop measures that do this. A small number of studies have separated them, including a thinking aloud study (M.J. Cooper & Fairburn, 1992a) and some Stroop studies. Interestingly, these studies are the only studies in the literature that show differences at a cognitive level between patients with anorexia nervosa and patients with bulimia nervosa. Vitousek cautions that hunger may be the most parsimonious explanation for differences between patients in such studies (Vitousek & Hollon, 1990; Vitousek, 1996) but this remains to be demonstrated empirically. While the core psychopathology may indeed reside in weight and shape assumptions this does not mean that eating and food related assumptions do not have an important role. As noted above, patient groups may differ in assumptions. For example, it is possible that different groups or clusters of assumptions may characterize different types of eating disorder: some types of assumption, perhaps reflecting a certain content area, may be relatively more important in one eating disorder than in another. The causal link More work needs to be done to investigate the causal role of underlying assumptions, as well as the role of core beliefs in contributing to the development of eating disorders. Improved assessment of assumptions,

Cognitive theory in eating disorders

139

using measures with adequate psychometric properties, separating out different areas of concern (weight-related from weight-related self-schemata) in the way suggested by Vitousek and Hollon, and separating weight and shape from food and eating concern, might facilitate understanding of how and which beliefs and assumptions are causally related to the development of eating disorders. Treatment studies might be used to identify specific predictors of relapse and prospective studies in the normal population might be conducted to identify which aspects of cognition or pattern of cognitive variables predict the development of an eating disorder. Formulation There is no overall, detailed cognitive theory of eating disorders that explains precisely how the different elements link together. At present the lack of a detailed theoretical model means that it is frequently difficult to derive adequate conceptualizations to guide treatment. Indeed, compared to other disorders, very little that is helpful to the clinician has been written on cognitive models of eating disorders. More detailed formulations are vital, both to explain the development of an eating disorder (a longitudinal formulation) and to explain the maintenance of characteristic symptoms (typical vicious circles), both in anorexia nervosa and in bulimia nervosa. A longitudinal formulation. The theory is not very detailed at this level. As noted above, it provides only a very general description of the content of beliefs (underlying assumptions and core beliefs) and, in particular, of how the different elements believed to be important might link together. For example, Vitousek and Hollon (1990) note that weight-related selfschemata are a combination of views about the self and views and information about weight and shape. However, they do not specify the mechanism, and, as noted above, the few examples they give of these beliefs appear to be at a superficial level (more typical of automatic thoughts than underlying meanings or beliefs). Mechanism is a key element in modelling dysfunction yet we know little about the relationship between the different elements, e.g. how one moves from core beliefs to self statements and disturbed behaviour, what the content and form of relevant intervening beliefs might be, how the different beliefs and assumptions identified, for example, by Vitousek and Hollon, relate to each other. Some of the suggestions for further research noted above might help clarify this but, in general, more very basic work is needed here: for example, a detailed semi-structured interview (such as that used to investigate core beliefs in health anxiety, Wells & Hackmann, 1993) might help to understand the nature of linking beliefs more clearly, including how they relate to or reflect schema driven processes. This type of interview might also help to identify the early

f

j ' •

140

M. Cooper

experiences that are important in the formation of assumptions and beliefs. I At present, no theory explains how patients arrive at the idea that dieting and losing weight will solve their problems and why they develop an eating disorder as opposed to some other kind of disorder. Vicious circle formulations. In other disorders, e.g. panic (D.M. Clark, 1986), vicious circle formulations have been developed to explain the maintenance of symptoms and are frequently used by clinicians to help design and target initial interventions. While existing theory notes that dieting maintains binge-eating, and that dichotomous thinking has a role in triggering specific episodes of binge-eating in bulimia nervosa, no detailed formulation, encompassing cognition, emotion and behaviour has been developed to explain both immediate triggers and the maintenance of dieting, bingeing and compensatory behaviours on a moment-by-moment basis, i.e. while the behaviours are actually occurring, in either anorexia nervosa or bulimia nervosa. Detailed, semi-structured interviews might help to clarify the precise nature of typical vicious circles that explain problematic behaviour in both disorders. This would be of particular help to clinicians trying to conceptualize the individual case early in treatment as well as provide a focus for initial interventions. Conclusion

This paper has reviewed four cognitive models of eating disorders, two that are commonly used to guide treatment and two that extend our understanding of the elements that need to be included in a comprehensive cognitive model. Evidence for a set of hypotheses derived from the models has been presented and discussed. It is concluded that initial evidence is promising and supports various aspects of cognitive theory. However much further research is needed. This should be designed to investigate the validity of existing theoretical contributions and to extend and develop theory. It seems likely that such research will be important for the future development of cognitive therapy in eating disorders. Further development of cognitive theory in eating disorders might help clinicians to conceptualize the individual case in more detail and to focus treatment on key cognitive issues. Ultimately, this should make treatment for bulimia nervosa more effective and facilitate its application to anorexia nervosa. References AGRAS, W.S., SCHNEIDER, J.A., ARNOW, B., RAEBURN, S.D., & TELCH, C.F. (1989).

Cognitive-behavioural and response prevention treatments for bulimia nervosa. Journal of Consulting and Clinical Psychology, 57, 215-222. AMERICAN PSYCHIATRIC ASSOCIATION (1994). Diagnostic and statistical manual of

Cognitive theory in eating disorders

141

mental disorders. Fourth Edition. Washington, D C : American Psychiatric Association. ANDERSON, J.R. (1983). The architecture of cognition. Harvard: Harvard University Press. BECK, A.T. (1976). Cognitive therapy and the emotional disorders. New York: International Universities Press. BECK, A.T. (1996). Beyond belief: a theory of modes, personality and psychopathology. In P.M. Salkovskis (Ed.), Frontiers of cognitive therapy (pp. 1-25). New York: Guilford Press. BECK, A.T., RUSH, A.J., SHAW, B.E, & EMERY, G. (1979). Cognitive therapy of

depression: A treatment manual. New York: Guilford Press. BEN-TOVIM, D.I., WALKER, M.K., FOK, D., & YAP, E. (1989). An adaptation of the

Stroop test for measuring shape and food concerns in eating disorders: A quantitative measure of psychopathology? International Journal of Eating Disorders, 8, 681-687. BEN-TOVIM, D.I. & WALKER, M.K. (1991). Further evidence for the Stroop test as a quantitative measure of psychopathology in eating disorders. International Journal of Eating Disorders, 5, 609-613. BLISS, E.L. & BRANCH, C.H. (1960). Anorexia nervosa: Its history, psychology and biology. New York: Hoeber. BRUCH, H. (1973). Eating disorders: Obesity, anorexia nervosa and the person within. New York: Basic Books. BUTLER, G. & MATHEWS, A. (1983). Cognitive processes in anxiety. Advances in Behaviour Therapy and Research, 5, 51-62. CHANNON, S. & HAYWARD, A. (1990). The effect of short-term fasting on processing of food cues in normal subjects. International Journal of Eating Disorders, 9, 447-452. CHANNON, S., HEMSLEY, D., & DE SILVA, P. (1988). Selective processing of food

words in anorexia nervosa. British Journal of Clinical Psychology, 27, 259-260. CHANNON, S., DE SILVA, P., HEMSLEY, D., & PERKINS, R. (1989). A controlled

trial of cognitive-behavioural and behavioural treatment of anorexia nervosa. Behaviour Research and Therapy, 27, 529-535. CLARK, D.A., FELDMAN, J., & CHANNON, S. (1989). Dysfunctional thinking in

anorexia and bulimia nervosa. Cognitive Therapy and Research, 13, 377-387. CLARK, D.A. & DE SILVA, W.P. (1985). The nature of depressive and anxious intrusive thoughts: Distinct or uniform phenomena? Behaviour Research and Therapy, 23, 383-393. CLARK, D.M. (1986). A cognitive approach to panic. Behaviour Research and Therapy, 24, 461-170. COLLINS, A.M. & LOFTUS, E.F. (1975). A spreading activation theory of semantic processing. Psychological Review, 82, 407-428. COOPER, M.J. (1991). Cognitive processes in anorexia nervosa and bulimia nervosa. Doctoral dissertation, University of Oxford. COOPER, M.J., ANASTASIADES, P., & FAIRBURN, C.G. (1992). Selective processing of

142

M. Cooper

eating, weight and shape related words in bulimia nervosa. Journal of Abnormal Psychology, 101, 352-355. COOPER, M.J., CLARK, D.M., & FAIRBURN, C.G. (1993). An experimental study

of the relationship between thoughts and eating behaviour in bulimia nervosa. Behaviour Research and Therapy, 31, 749-757. COOPER, M.J. & FAIRBURN, C.G. (1992a). Thoughts about eating, weight and shape in anorexia nervosa and bulimia nervosa. Behaviour Research and Therapy, 30, 501-511. COOPER, M.J. & FAIRBURN, C.G. (1992b). Selective processing of eating, weight and shape related words in patients with eating disorders and dieters. British Journal of Clinical Psychology, 31, 363-365. COOPER, M.J. & FAIRBURN, C.G. (1993). Demographic and clinical correlates of selective information processing in patients with bulimia nervosa. International Journal of Eating Disorders, 13, 109-116.

\

; \

COOPER, M.J., TODD, G., & COHEN-TOVEE, E. (1996). Core beliefs in eating dis-

orders. International Cognitive Therapy Newsletter, 10, part 2, 2-3. COOPER, P.J. & FAIRBURN, C.G. (1984). Cognitive-behaviour therapy for anorexia nervosa: Some preliminary findings. Journal of Psychosomatic Research, 28, 493-500. COOPER, P.J. & STEERE, J. (1995). A comparison of two psychological treatments for bulimia nervosa: Implications for models of maintenance. Behaviour Research and Therapy, 33, 875-885. COOPER, Z. & FAIRBURN, C.G. (1987). The Eating Disorder Examination: A semistructured interview for the assessment of the specific psychopathology of eating disorders. International Journal of Eating Disorders, 6, 1-8. DERRY, P.A. & KUIPER, N.A. (1981). Schematic processing and self-reference in clinical depression. Journal of Abnormal Psychology, 90, 286-297. DRITSCHEL, B.H., WILLIAMS, K., & COOPER, P.J. (1991). Cognitive distortions

amongst women experiencing bulimic episodes. International Journal of Eating Disorders, 10, 547-555. FAIRBURN, C.G. (1981). A cognitive-behavioural approach to the treatment of bulimia nervosa. Psychological Medicine, 11, 707-711.

J

FAIRBURN, C.G., COOPER, P.J., COOPER, M.J., MCKENNA, F.P., & ANASTASIADES, P.

(1991). Selective information processing in bulimia nervosa. International Journal of Eating Disorders, 10, 415-422.

'.

FAIRBURN, C.G., COOPER, Z., & COOPER, P.J. (1986) The clinical features and

maintenance of bulimia nervosa. In K.D. Brownell and J.P. Foreyt (Eds.), Physiology, psychology and treatment of the eating disorders. New York: Basic Books. FAIRBURN, C.G., JONES, R., PEVELER, R.C., CARR, S.J., SOLOMON, R.A., O ' C O N N O R ,

M.E., BURTON, J., & HOPE, R.A. (1991) Three psychological treatments for bulimia nervosa: A comparative trial. Archives of General Psychiatry, 48, 463-469.

j

FAIRBURN, C.G., KIRK, J., O ' C O N N O R , M., & COOPER, P.J. (1986). A comparison

\

of two psychological treatments for bulimia nervosa. Behaviour Research and Therapy, 24, 629-643.

Cognitive theory in eating disorders

143

FAIRBURN, C.G., NORMAN, P.A., WELCH, S.L., O ' C O N N O R , M.E., D O L L , H.A., &

PEVELER, R.C. (1995). A prospective study of outcome in bulimia nervosa and the long-term effects of three psychological treatments. Archives of General Psychiatry, 52, 304-312. FAIRBURN, C.G., PEVELER, R . C , JONES, R., H O P E , R.A., & D O L L , H.A. (1993).

Predictors of 12-month outcome in bulimia nervosa and the influence of attitudes to shape and weight. Journal of Consulting and Clinical Psychology, 61, 696-698. FISCHLER, I. & BLOOM, P.A. (1979). Automatic and attentional processes in the effects of sentence contents on word recognition. Journal of Verbal Learning and Verbal Behaviour, 18, 1-20. FRANKO, D.L. & ZUROFF, D.C. (1992). The Bulimic Automatic Thoughts Test: Initial reliability and validity data. Journal of Clinical Psychology, 48, 505-509. FREEMAN, C.P.L., BARRY, F., DUNKELD-TURNBULL, J., & HENDERSON, A. (1988).

Controlled trial of psychotherapy for bulimia nervosa. British Medical Journal, 296, 521-525. GARNER, D.M. & BEMIS, K.M. (1982). A cognitive-behavioural approach to anorexia nervosa. Cognitive Therapy and Research, 6, 123-150. GARNER, D.M. & GARFINKEL, P.E. (1979). The Eating Attitudes Test: An index of the symptoms of anorexia nervosa. Psychological Medicine, 9, 273-279. GILLIGAN, S.G. & BOWER, G.H. (1984). Cognitive consequences of emotional arousal. In C. Izard, J. Kagan and R. Zajonc (Eds.), Emotions, cognition and behaviour. New York: Cambridge University Press. GREEN, M.W., MCKENNA, F.P., & DE SILVA, M.S.L. (1994). Habituation patterns to

colour naming of eating-related stimuli in anorexics and non-clinical controls. British Journal of Clinical Psychology, 33, 499-509. GUIDANO, V.F. & LIOTTI, G. (1983). Cognitive processes and emotional disorders: A structural approach to psychotherapy. New York: Guilford Press. KING, G.A., POLIVY, J., & HERMAN, C.P. (1991). Cognitive aspects of dietary

restraint: Effects on person memory. International Journal of Eating Disorders, 10, 313-321. KIRKLEY, B.G., SCHNEIDER, J.A., AGRAS, W.S., & BACHMAN, J.A. (1985). A com-

parison of two group treatments for bulimia nervosa. Journal of Consulting and Clinical Psychology, 53, 4 3 ^ 8 . LANDAU, R.J. (1980). The role of semantic schemata in phobic word recognition. Cognitive Therapy and Research, 4, 427—434. LANDAU, R.J. & GOLDFRIED, M.R. (1981). The assessment of schemata: A unifying framework for cognitive, behavioural and traditional assessment. In P.C. Kendall and S.D. Hollon (Eds.), Assessment strategies for cognitive-behavioural interventions. New York: Academic Press. LEITENBERG, H., ROSEN, J.C., GROSS, J., NUDELMAN, S., & VARA, L.S. (1988).

Exposure plus response prevention of bulimia nervosa: A controlled evaluation. Journal of Consulting and Clinical Psychology, 56, 535-541. LONG.C.G., HINTON, C , & GILLESPIE, N.K. (1994). Selective processing of food

144

M. Cooper

and body size words: Application of the Stroop test with obese restrained eaters, anorexics, and normals. International Journal of Eating Disorders, 15, 279-283. MARKUS, H., HAMILL, R, & SENTIS, K.P. (1987). Thinking fat: Self schemata for

body weight and the processing of weight relevant information. Journal of Applied Social Psychology, 17, 50-72. MATHEWS, A., RICHARDS, A., & EYSENCK, M. (1989). Interpretation of homophones

related to threat in anxiety states. Journal of Abnormal Psychology, 98, 31-34. MITCHELL, J.E. & PYLE, R.L. (1982). The bulimic syndrome in normal weight individuals: A review. International Journal of Eating Disorders, 1, 61-73. MIZES, J.S. (1988). Personality characteristics of bulimic and noneating disordered female controls: A cognitive-behavioural perspective. International Journal of Eating Disorders, 7, 541-550. MIZES, J.S. (1990). Criterion-related validity of the Anorectic Cognitions Questionnaire. Addictive Behaviours, 15, 153-163. MIZES, J.S. (1991). Construct validity and factor stability of the Anorectic Cognitions Questionnaire. Addictive Behaviours, 16, 89-93. MIZES, J.S. (1992). Validity of the Mizes Anorectic Cognitions Scale: A comparison between anorectics, bulimics and psychiatric controls. Addictive Behaviours, 17, 283-289. MIZES, J.S. & CHRISTIANO, B.A. (1995). Assessment of cognitive variables relevant to cognitive behavioural perspectives on anorexia nervosa and bulimia nervosa. Behaviour Research and Therapy, 33, 95-106. MIZES, J.S. & KLESGES, R.C. (1989). Validity, reliability, and factor structure of the Anorectic Cognitions Questionnaire. Addictive Behaviours, 14, 589-594. MOGG, K., MATHEWS, A., & WEINMAN, J. (1987). Memory bias in clinical anxiety. Journal of Abnormal Psychology, 96, 94-98. PADESKY, C.A. & GREENBERGER, D. (1995). Clinicians guide to mind over mood. New York: Guilford Press. PADESKY, C.A. & MOONEY, K.A. (1993). Cognitive therapy of personality disorders: Schema change processes. Workshop presented at the European Congress of Behaviour and Cognitive Therapy, London, September. PERPINA, C , HEMSLEY, D., TREASURE, J., & DE SILVA, P. (1993). Is the selective

information processing of food and body words specific to patients with eating disorders? International Journal of Eating Disorders, 14, 359-366. PERSONS, J.B. (1989). Cognitive therapy in practice: A case formulation approach. New York: Norton & Company. PHELAN, P.W. (1987). Cognitive correlates of bulimia: The Bulimic Thoughts Questionnaire. International Journal of Eating Disorders, 6, 593-607. POWERS, PS, SCHULMAN, R.G., GLEGHORN, A.A., & PRANGE, M.E. (1987). Percep-

tual and cognitive abnormalities in bulimia. American Journal of Psychiatry, 144, 1456-1460. REYNAMCGLONE, C.L., OLLENDICK, T.H., & HART, K.J. (1986). Self-efficacy and

bulimia: Development and validation of the Self-Assessment Scale. Poster pre-

Cognitive theory in eating disorders

145

sented at the Association for the Advancement of Behaviour Therapy, Chicago, November. RUSSELL, G.F.M. (1979). Bulimia nervosa: An ominous variant of anorexia nervosa. Psychological Medicine, 9, 429-448. ROYAL COLLEGE OF PSYCHIATRISTS (1992). Eating disorders. Council Report CR14. London: Royal College of Psychiatrists. SCANLON, E., OLLENDICK,, T.H., & BAYER, K. (1986). The role of cognitions in

bulimia: An empirical test of basic assumptions. Paper presented at the Association for the Advancement of Behaviour Therapy, Chicago, November. SCHOTTE, D.E., MCNALLY, R.J., & TURNER, M.L. (1990). A dichotic listening

analysis of body weight concern in bulimia nervosa. International Journal of Eating Disorders, 9, 109-113. SCHULMAN, R.G.,

KINDER, B.N.,

POWERS, P.S., PRANGE, M., & GLEGHORN, A.

(1986). The development of a scale to measure distortions in bulimia. Journal of Personality Assessment, 50, 630-639. SEBASTIAN, S.B., WILLIAMSON, D.A., & BLOUIN, D.C. (1996). Memory bias for

fatness stimuli in the eating disorders. Cognitive Therapy and Research, 20, 275-286. SELVINI-PALAZZOLI, M.P. (1974). Self-starvation London: Chaucer. STROOP, P.J. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643-661. THACKWRAY, D.E., SMITH, M.C., BODFISH, J.W., & MEYERS, A.W. (1993). A com-

parison of behavioural and cognitive-behavioural interventions for bulimia nervosa. Journal of Consulting and Clinical Psychology, 61, 639-645. THOMPSON, D.A., BERG, K.M., & SHATFORD, L.A. (1987). The heterogeneity of

bulimia symptomatology: Cognitive and behavioural dimensions. International Journal of Eating Disorders, 6, 215-234. VITOUSEK, K.B. & HOLLON, S.D. (1990). The investigation of schematic content and processing in the eating disorders. Cognitive Therapy and Research, 14, 191-214. VITOUSEK, K.M. (1996). The current status of cognitive-behavioural models of anorexia nervosa and bulimia nervosa. In P.M. Salkovskis (Ed.), Frontiers of cognitive therapy (pp. 383-418). New York: Guilford Press. WELLS, A. & HACKMANN, A. (1993). Imagery and core beliefs in health anxiety: Content and origins. Behavioural and Cognitive Psychotherapy, 21, MA—ITS. YOUNG, J.E. (1990). Cognitive therapy for personality disorders: A schema focused approach. Sarasota, FL: Professional Resource Exchange. ZOTTER, D.L. & CROWTHER, J.H. (1991). The role of cognitions in bulimia nervosa. Cognitive Therapy and Research, 5, 413—426.