Clinical Psychology and Psychotherapy Clin. Psychol. Psychother. 20, 319–333 (2013) Published online 20 November 2012 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/cpp.1823
Emotion Coupling and Regulation in Anorexia Nervosa John R. E. Fox,1,2,3* Emily Smithson,1 Sarah Baillie,1 Nuno Ferreira,3 Ingrid Mayr1 and Michael J. Power3 1
Eating Disorders Unit, Russell House, Priory Hospital Cheadle Royal, Cheadle, Cheshire UK Division of Clinical Psychology, University of Manchester, Manchester, UK 3 Clinical Psychology, School of Health in Social Science, University of Edinburgh, Medical School, Edinburgh, UK 2
Objective: The present study sought to investigate emotion regulation strategies in people with anorexia nervosa (AN) and whether the theoretical concept of ‘emotion coupling’ between anger and disgust could help to explain some of the specific eating disorder symptomatology in people with AN. Method: This ‘emotion coupling’ hypothesis was tested using a mood induction procedure within laboratory conditions, where individuals with AN (n = 22) were matched with control participants (n = 19). Participants completed a bank of different measures prior to the study, and these included measures of eating pathology, core beliefs about the self and others, and emotion regulation strategies. Within the experimental part of this study, anger, disgust and body size estimation were measured prior to and after an anger induction procedure (i.e., a repeated measures design). Results: People with AN demonstrated a significantly more internal-dysfunctional way to regulate their emotional states, when compared with matched controls. Within the ‘emotional coupling’ part of the study, participants showed a significant increase in levels of disgust and body size estimation following an anger induction when compared with matched controls. Discussion: The significance of these results was considered in the light of the new Schematic, Propositional, Analogical and Associative Representation Systems in eating disorders model. Copyright © 2012 John Wiley & Sons, Ltd. Key Practitioner Message • Emotions and emotion processing are problematic for people with anorexia nervosa. • Overestimation in body size for people with anorexia nervosa may be due to the coupling of anger and disgust. • Clinicians need to consider the importance of automatic cognitive routes of emotion elicitation that may maintain certain eating disorder symptoms, such as poor body image. Keywords: anorexia nervosa, basic emotions, emotion regulation, anger, disgust
INTRODUCTION Across the literature, it has been argued that our understanding of anorexia nervosa (AN) is still poor (Cooper, 2005; Fox & Power, 2009). Despite this, a number of authors have proposed some theoretical accounts of the emotional processes in AN (e.g., Hatch et al., 2010; Haynos & Fruzzetti, 2011). These theories highlight that there are a number of emotion regulation difficulties in AN, and these difficulties occur at an unconscious level within milliseconds of the triggering event. Other authors have also discussed how various eating disorder symptoms may have the function of either suppressing or blocking painful emotions (e.g., Brockmeyer et al., 2012; *Correspondence to: Dr John Fox, Clinical Psychology, School of Health in Social Science, University of Edinburgh, Medical School, Edinburgh EH8 9AG, UK. E-mail:
[email protected]
Copyright © 2012 John Wiley & Sons, Ltd.
Waller, Kennerley, & Ohanian, 2007; Serpell, Treasure, Teasdale, & Sullivan, 1999; Wonderlich, Mitchell, Peterson, & Crow, 2001). For example, Brockmeyer et al. (2012) reported data that showed that low weight may serve as a dysfunctional behaviour in order to regulate painful emotions in AN. Whilst for more bulimic behaviours, Cooper, Wells, and Todd (2004, 2004) proposed that bingeing and vomiting are often used when the self is faced with overwhelming emotion, because these symptoms can reduce the physiological experience of negative affect.
EMOTION PROCESSES IN ANOREXIA NERVOSA Although the idea of eating disorder symptomatology being used as an emotion regulation strategy is now a fairly established theoretical process in the understanding
J. R. E. Fox et al.
320 of eating disorders (e.g., Bydlowski et al., 2005), little is still known about the specifics of these processes (e.g., Fox & Power, 2009; Davies et al., in press). For example, given the extent of theory and research on normal emotion functioning, the role of individual emotions and their relation to each other is still relatively unknown in AN. However, there are some indications from the research literature as to which emotions may be especially important in AN. Geller, Cockell, Hewitt, Goldner, and Flett (2000) found that people with AN suppressed anger more than controls (as measured by the State-Trait Anger Expression Inventory [STAXI]; Spielberger, 1996), and this suppression predicted body dissatisfaction. The authors of this study argued that people with AN suppressed anger because they were fearful to assert themselves with others as they were scared that they would be rejected, and they named this process ‘silencing the self’. Moreover, Waller et al. (2003) reported data from the STAXI that showed how an eating disorder participant group had higher state anger and anger suppression scores when compared with controls. Hayaki, Friedman, and Brownell (2002) found evidence to show that body dissatisfaction was predicted by poor emotion expression when depression and assertiveness were controlled for in their analysis. Data from these studies have highlighted the usefulness of self-report measures of anger and emotional expression, because the results from this research have shown evidence that when anger is suppressed, it may lead to increased body dissatisfaction. In a series of recent studies, Fox and colleagues have explored the relationship between basic emotions and disordered eating (e.g., Fox & Harrison, 2008; Fox, 2009; Ioannou & Fox, 2009; Fox & Froom, 2009). Fox (2009) found in a grounded theory study of AN that the emotions of anger and, to a lesser degree, sadness were discussed as being ‘toxic’ or ‘shaming’ and were inhibited. The participants discussed how anger seemed to be particularly toxic, and this ‘toxicity’ was pivotal in the rise of eating disorder symptoms. However, there were differences in individual beliefs about the basic emotions and why they should not be expressed. Anger was regarded as being ‘dangerous’ and ‘harmful’ to both the participant and other people, whilst for sadness, its expression was a perceived sign of ‘weakness’. The analysis highlighted how people with AN often did not regard themselves as entitled to be an ‘emotional’ human being, and the expression of emotion was seen to lead to rejection from significant others (in keeping with Geller et al., 2000).
THE SPAARS-ED MODEL AND ‘EMOTIONAL COUPLING’ IN ANOREXIA NERVOSA Fox and Power (2009) developed the SPAARS-ED model from the general SPAARS model of emotion and emotional Copyright © 2012 John Wiley & Sons, Ltd.
disorder (Power & Dalgleish, 2008). This acronym refers to the individual components of the model, which are Schematic, Propositional, Analogical and Associative Representation Systems in eating disorders.1 Across the SPAARSED model, it is argued that eating disorders are driven by both an avoidance of emotion, via restriction and/ or bingeing/vomiting, and the directing of painful emotion onto the body, in the form of self-disgust/shame (see also Goss & Allan, 2009). This redirection of emotion leaves the self with a strong propensity to feel revulsion to both body and food in AN (Figure 1). According to the literature, elicitation of emotion appears to be driven both by intrapersonal and interpersonal factors (e.g., Oatley & Johnson-Laird, 1987), which are applicable to AN. Research has consistently highlighted that people with AN have marked social difficulties (e.g., Treasure, Corfield, & Cardi, in press), and these difficulties seem to be connected to beliefs about the self being at risk of being rejected and being worthless or bad in some way (e.g., Cooper, 1997; Cooper & Hunt, 1998; Leung, Waller, & Thomas, 1999; Waller, Ohanian, Meyer, & Osman, 2000). In the standard cognitive behavioural therapy (CBT) literature, these beliefs could be formulated as core beliefs, but the SPAARS-ED model takes these beliefs a step further by regarding them as a more general process of cognition, which would include different forms of representations (e.g., visual, tactile and verbal) and, therefore, above more ‘verbal expressible propositional concepts’ (Jones, 2001). This schematic processing of information captures a lot more of the complexity that is present in clinical presentations, such as intellectually knowing that one is not ‘fat’ but still ‘feeling fat and disgusting’. Fox and Power (2009) argued that the appraisal systems work within three domains of knowledge, namely the self, others and the World (in keeping with other cognitive authors, such as Beck, Rush, Shaw, & Emery, 1979). Due to these complex cognitive processes, it is hard to capture the full meaning of these processes in language. However, these appraisals could be understood as ‘SELF-BAD, OTHERS-IMPORTANT and WORLD-UNSAFE’, and the resulting emotions would be sadness and disgust (in keeping with the findings of Troop, Treasure, & Serpell, 2002; Power & Tarsia, 2007). In other words, the self is regarded as being ‘bad’ or ‘worthless’ and at high risk of being rejected by others. This places other people into an important position, where their needs are regarded as being more important, and they need to be appeased and not upset. The fear here is that the other person will find out what they are ‘really like’ (e.g., bad/ contemptible) and reject them. This also leaves the World feeling unsafe and unpredictable. 1
It is beyond the scope of this paper to discuss this model in detail, but we would like to refer the reader to both Power and Dalgleish (2008) and Fox and Power (2009) for a fuller consideration of the general and eating disorder-specific model.
Clin. Psychol. Psychother. 20, 319–333 (2013)
Emotion Coupling and Regulation in Anorexia Nervosa
Bodily sensations clothes Food Visual Images Verbal statements Smells (e.g. of food)
Associative route Internal attribution towards body (e.g. fat) External attribution towards food
Propositional Level ‘I am worthless’ ‘people always reject me’ ‘my emotions scare me’ ‘I am fat’ ‘I hate food’/ ‘I must avoid food’ ‘I am a disgusting individual’
ANGER DISGUST SADNESS
DISGUST (Self)
Eating disorder symptomatology (effects of starvation – feedback back into the emotional processing)
Interpersonal event
Schematic models SELF-BAD (e.g. self is worthless) OTHERS – IMPORTANT (e.g. fear of rejection) WORLD – UNSAFE (e.g. people hurt me) ANGER –WEAK ANGER – DANGEROUS FEAR – DISGUST SELF-SHAME
COUP LING
Analogue system
321
Figure 1. The SPAARS model of eating disorders (SPAARS-ED)
Research by Geller et al. (2000), Waller et al. (2003), Fox and Harrison (2008), Fox (2009) and Fox and Froom (2009) highlighted the emotion of anger as a highly distressing emotion for people with AN. According to the SPAARS-ED model, anger may be appraised as being ego-dystonic and thus detached from the person’s sense of self (for example, ANGER-DANGEROUS; ANGERTHREAT-OTHERS). This ego-dystonic anger is suppressed, via restriction/bingeing–vomiting. The findings from Fox (2009) highlighted how people with AN may perceive themselves as not being entitled to feel emotion, and expressing anger is seen as risking rejection from others. It has long been noted that AN is often associated with poor social skills (Treasure et al., in press), and these difficulties, it is argued, are pivotal in the genesis of AN (Schmidt & Treasure, 2006). Research and theory have discussed how losing weight is often seen as a means to be more respected and, hence, likeable and to feel more in control (e.g., Fox, Federici, & Power, 2012), and the parallel processes of emotion suppression and weight loss become connected over time. Thus, the SPAARS-ED model argues that the suppression of anger in AN becomes ‘coupled’ with the emotion of disgust. This theoretical process has received some empirical support, in that Fox and Harrison (2008) found that when they induced anger in participants with disordered eating patterns, they showed significantly higher levels of anger and disgust following anger induction when compared with controls. Furthermore, they also reported significantly more suppression/non-expression of anger. Although this study only tested this hypothesis on participants who were students with elevated levels of disordered eating, the results were striking. These findings and those of Geller et al. (2000) and Hayaki et al. (2002) also show some preliminary evidence that the lack of expression of emotions, in particular anger, is related to disgust and body dissatisfaction. Copyright © 2012 John Wiley & Sons, Ltd.
EXPERIMENTAL DESIGNS IN EMOTIONS AND EATING DISORDERS RESEARCH Most of the research that has highlighted the role of emotions in eating disorders has used either non-experimental or qualitative methodologies. When designing the current study, it was decided to use a two-stage design that incorporated both cross-sectional and experimental components. This decision was taken because although the previous research has produced some valuable findings, it is still limited due to the lack of precision in knowing the impact of a specific emotion on an eating disorder symptom (e.g., increased body size estimation). However, some authors have developed experimental techniques, and it was envisaged that a review of this work would allow for the development of the methodology for this current study. Davies, Schmidt, Stahl, and Tchanturia (2011), Warren, Strauss, Taska, and Sullivan (2005) and Harrison, Taylor, and Marske (2006) have demonstrated that experimental techniques can be used to induce emotion, mainly negative emotion, and then measure the effect on the specific eating pathology under investigation. For example, Warren et al. (2005) found that people who restrained their eating were more likely to eat more snacks following exposure to a sad movie and an increase in diet commercials, compared with neutral commercials. Likewise, Davies et al. (2011) demonstrated that people with AN significantly avoided attending to sad film clips more than healthy controls. What is interesting about all these studies is that they have used the easiest negative emotion to induce, namely sadness. Anger was conspicuous in its absence, and this seemed to be due to both conceptual limitations (e.g., lumping all negative emotion together) and practical problems of inducing an interpersonal emotion in a controlled, experimental way. In the more general emotions research literature, Velten addressed this point by developing a technique that used Clin. Psychol. Psychother. 20, 319–333 (2013)
J. R. E. Fox et al.
322 the repeating of words and phrases to induce an emotion (Velten, 1968). However, this technique has some serious ecological limitations, and it was decided in the development of this research study’s methodology to induce emotion by using the Velten technique of rehearsal as a guiding principle.
AIMS/HYPOTHESES Aim As shown in the Introduction, research has begun to show that emotions are of central importance in understanding AN. The SPAARS-ED model makes a number of predictions about how individuals with AN may have negative core beliefs and an emotion regulation style that prevents the expression of emotion. It is argued that these negative core beliefs and internal-dysfunctional emotion regulation strategies (e.g., suppression of emotion, self-harm etc.) create an internal world where negative emotions are difficult to experience and are managed via emotion coupling. However, what is not known is if this theoretical idea of emotion coupling occurs within individuals with a diagnosis of AN, as well as whether the induction of anger has a direct effect upon levels of both disgust and body shape estimation (as reported by Fox & Harrison, 2008). These theoretical questions underpin the research reported in this paper. Hypothesis 1 This is a replication hypothesis because previous research has identified that people with AN have significant difficulties with emotion regulation (e.g., Gilboa-Schechtman, Avnon, Zubery, & Jeczmien, 2006; Harrison, Tchanturia, & Treasure, 2010). Therefore, it was hypothesized that AN participants would score significantly higher on the internal-dysfunctional subscales and significantly lower on the internal-functional subscales of the Regulation of Emotion Questionnaire (REQ; Phillips & Power, 2007) than control groups. It was also hypothesized that the AN participants would score significantly higher on the external-dysfunctional subscales and significantly lower on the external-functional subscales of the REQ than the control group. It was also hypothesized that there would be a significant difference between AN and the control participants on the negative self scale of the Brief Core Schema Scale (BCSS; Fowler et al., 2006). Hypothesis 2 To investigate the relationships between the Basic Emotions Scale (BES; Power, 2006) and the BCSS (Fowler et al., 2006) Copyright © 2012 John Wiley & Sons, Ltd.
for the AN group. It was hypothesized that there would be a significant positive correlation between negative emotion (i.e., anger, sadness and disgust) and the negative self and negative other beliefs. Hypothesis 3 On the basis of the findings of Fox and Harrison (2008), AN participants would report a significant increase in state anger (as measured by the State-Trait Measure of Anger [STAXI]; Spielberger, 1996), disgust (as measured by the Disgust Scale [DS-R]; Olatunji et al., 2007) and the estimation of body size (as measured by body shape silhouettes [BSS]) when compared with matched controls following an anger induction. Hypothesis 4 It was predicted that estimation of body size (as measured by BSS) would significantly positively correlate with shape concern, weight concern subscales (from the Eating Disorder Examination Questionnaire [EDE-Q]; Fairburn & Beglin, 1994), state disgust (as measured by the BES; Power, 2006) and estimation of body size (as measured by BSS) for the AN group. For all of the hypotheses, it was predicted that anxiety and depression as measured by the Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, 1983) would covary in the analyses (in keeping with analyses undertaken by Fox & Harrison, 2008).
METHODOLOGY Design This study utilized a mixed design that incorporated both a between and a within participants design. For the sake of clarity, the results will be presented in two stages, with the between subjects design (cross-sectional/questionnaire) being presented first and the mixed design being presented second.
Participants Across the study, only female participants were recruited. For the AN group, participants were recruited from three clinical services in the north of England and Scotland, whilst control participants were recruited from university students in the north of England. The clinical services that assisted in the data collection were regional eating disorders centres. These are specialist services for people with severe AN and bulimia nervosa. Diagnoses were Clin. Psychol. Psychother. 20, 319–333 (2013)
Emotion Coupling and Regulation in Anorexia Nervosa confirmed by multiple clinicians who have a significant amount of experience of working with and treating eating disorders, and their mean number of years of practice in working with people with eating disorders was 7 years. Diagnoses made adhered to the criteria provided by DSM-IV (APA, 1994). Furthermore, by the time people were referred to the eating disorders services used in this study, they would have had multiple assessments, and their diagnosis would have been confirmed by multiple clinicians. However, in order to double check the clinical caseness of all the participants, body mass index (BMI) data were collected, and data were also collected with the EDE-Q (Fairburn & Beglin, 1994). BMI was directly collected from clinical notes for the AN group, but this was not possible for the control group. Therefore, the control group provided their own weights and heights. The EDE-Q was selected because it is a quick and reliable measure of eating disorder symptoms and there are published norms for this scale (see Mond, Hay, Rodgers, Owen, & Beumont, 2004). However, the EDE-Q asks questions about current levels of eating, bingeing, vomiting and exercise levels, and due to the inpatient units having a strong and set supervised eating programme (i.e., meals are at set times and are non-negotiable), scores on the items pertaining to eating and vomiting would be artificially reduced. This would mean that the actual level of symptomatology would potentially be under-reported on the EDE-Q. Fairburn and Beglin (1994) stated that a score of four on any of the subscales represented clinical caseness. In another study on the EDE-Q (Mond et al., 2004), it was found that a score of 3.09 on the global scale of the EDE-Q indicated clinical caseness of eating pathology. On the basis of these data, a cut-off of 3.09 was used to ensure clinical caseness for the AN group (due to the limitations of using the EDE-Q with an inpatient participant group). AN participants were only included in the analysis if their EDE-Q scores were in the clinical range and they had a clinical diagnosis of AN and a BMI lower than 17.5. As detailed in the procedure, this study had two parts, and it was possible for participants to consent to take part in the first part of the study but decide not to take part in the second part. This meant that 32 AN participants took part in the first part of the study (i.e., the collection of data via questionnaires) and 22 AN participants took part in the second part of the study (the experimental part of the study). It was decided to use a student comparison group because it was observed that many of the patients on the units from where the data were collected had been or were currently students and that they shared similar educational and social economic status. In order to recruit control participants, an administrator emailed a short invitation to take part in the study to students. This short invitation gave a brief summary of what the study was about by saying that it was a study designed to look at emotions and emotional Copyright © 2012 John Wiley & Sons, Ltd.
323 processing in people with AN. It explained that the study was in two parts, with the first part being undertaken on the website at Stellar Survey and the second part would be undertaken at a venue at the University where participants would be seen individually. The EDE-Q was used to ensure that the participants did not have an identifiable eating disorder. Mond et al. (2004) reported that a mean EDE-Q score of 1.19 indicated no eating pathology. In order to ensure that the control group did not have any eating pathology, a cut-off score of 1 was used, and their BMI had to be above 20. This cut-off was decided on because it ensured that people within the control group had a BMI that was significantly higher than the underweight bracket (
