A Risk and Maintenance Model for Bulimia Nervosa: From Impulsive Action to Compulsive Behavior

Abstract

This paper offers a new model for bulimia nervosa (BN) that explains both the initial impulsive nature of binge eating and purging as well as the compulsive quality of the fully developed disorder. The model is based on a review of advances in research on BN and advances in relevant basic psychological science. It integrates transdiagnostic personality risk, eating disorder specific risk, reinforcement theory, cognitive neuroscience, and theory drawn from the drug addiction literature. We identify both a state-based and a trait-based risk pathway, and we then propose possible state-by-trait interaction risk processes. The state-based pathway emphasizes depletion of self-control. The trait-based pathway emphasizes transactions between the trait of negative urgency (the tendency to act rashly when distressed) and high-risk psychosocial learning. We then describe a process by which initially impulsive BN behaviors become compulsive over time, and we consider the clinical implications of our model.

Introduction

This paper presents a theory of risk and maintenance for bulimia nervosa (BN) that explains both the initial, impulsive engagement in binge eating and purging and subsequent compulsive engagement in those behaviors, despite the harms they bring. To explain the model and its existing empirical support, we review selected clinical science literature on BN and basic science literature that sheds light on both initial engagement in bulimic behaviors and the compulsive quality of the full-threshold disorder.

BN is a disorder characterized by recurrent episodes of binge eating (consuming an unambiguously large amount of food in a discrete period of time and feeling unable to stop) followed by recurrent inappropriate compensatory behavior in an effort to prevent weight gain (such as self-induced vomiting: American Psychiatric Association (APA), 2013). BN is also characterized by an overvaluation of shape and weight, which influences self-evaluation (APA, 2013). Bingeing and purging behavior tends to be more prevalent in females (Hoek, 2006; Striegel-Moore & Bulik, 2007): It is estimated that 0.5% of men compared with 1.5% of women are diagnosed with BN in a lifetime (Kessler, Berglund, Chiu, Demler, Heeringa, et al., 2004; Kessler, Chiu, Dernier, & Walters, 2005).

In order to understand risk for the onset of bulimic behaviors and the development of compulsive engagement in those behaviors, several categories of factors must be considered and integrated. Both aspects of the environment and characteristics of the person are relevant, as is the interaction of the two. Broad, transdiagnostic risk factors appear to play important roles in conjunction with eating disorder-specific risk processes. The impact of momentary events must also be incorporated with stable, dispositional risks. We offer one such integrative model.

Consistently, researchers have reported that a portion of pre-adolescent or early adolescent girls engage in binge eating and purging behavior (Beato-Fernandez, Rodriguez-Cano, Belmonte-Llario & Martinez-Delgado, 2004; Bryant-Waugh & Lask, 1995; Combs, Spillane, Caudill, Stark, & Smith, 2012; Cotrufo, Cella, Cremato, & Labella, 2007; Franko & Omori, 1999; Gardner, Stark, Friedman, & Jackson, 2000; Shisslak, Crago, McKnight, Estes, Gray, & Parnaby, 1998; Shomaker, Tanofsky-Kraff, Elliott, Wolkoff, Columbo, et al., 2010; Tanofsky-Kraff, Shomaker, Olsen, Rozan, Wolkoff, et al., 2011). In one study, between 8.4% and 12.1% of girls ages 11–12 reported engaging in at least one binge eating episode in the preceding two weeks (Combs et al., 2012), with other reports indicating higher rates for early adolescent girls (Shomaker et al., 2010). In addition, between 2.6% and 4.8% of girls reported at least one purging episode in the 11–12 age range (Combs et al., 2012).

This early experience of symptomatic behavior is striking in its own right, and is also important because eating disorder symptoms and risk behaviors in early adolescence are highly predictive of later, diagnosable disorders (Killen, Taylor, Hayward, Wilson, Haydel et al., 1994; Kotler, Cohen, Davies, Pine, & Walsh, 2001; Smith, Simmons, Flory, Annus, & Hill, 2007). Anorexia nervosa (AN) and BN symptoms at the beginning of adolescence correlate greater than r = .40 with symptoms during adulthood, and diagnosable BN at the beginning of adolescence is associated with a 9-fold increase in BN and a 20-fold increase in AN during late adolescence (Kotler et al., 2001).

Even with treatment, bulimic pathology often becomes chronic. Only about 40% of individuals treated for BN experience lasting remission and the more severe the symptoms, the lower the recovery rates (Stice & Bulik, 2008). This is true even when individuals are treated with cognitive behavioral therapy, the most empirically supported treatment available for BN (Bohon, Stice, & Burton, 2009; Fairburn, Marcus, & Wilson, 1993; Wilson, 2005). The chronicity of the disorder is striking in part because engaging in bulimic behaviors significantly increases the risk of experiencing a number of health problems, including enlarged salivary glands, significant and permanent loss of dental enamel, esophageal tears, gastric rupture, cardiac arrhythmias, obesity, high blood pressure, type 2 diabetes, and production of fluid and electrolyte abnormalities (APA, 2013; Bulik, Sullivan, & Kendler, 2002). Furthermore, the life disruption associated with BN is dramatic and includes depression, anxiety disorders and substance use problems (APA, 2013; Bulik et al., 2002). A crucial question is what accounts for the disorder’s chronicity, given its profound negative consequences: Why do women with BN continue to engage in the behaviors even as they experience the harms?

Before introducing our model, it is important to note the limitation that the existing BN literature focuses very heavily on European American women. Given the focus on women, we use female terms, such as “women” and “girls” in presenting our model. In the closing of this paper, we discuss the available research conducted across ethnic groups and how these findings can be applied to our risk and maintenance theory for BN. We hope future research will test the suitability of the model we describe, and other models, to men and to women of diverse backgrounds.

Existing Risk Models for BN

To effectively summarize existing risk models for BN, it is necessary to briefly review and define the key factors in those models. Readers interested in more detailed reviews of specific risk factors can consult Jacobi, Hayward, de Zwaan, Kraemer, and Agras (2004), Pearson, Riley, Davis, and Smith (2014a), and Stice (2002). Relevant risk processes include those that are transdiagnostic, in that they confer risk for multiple forms of dysfunction, and those that are eating disorder-specific. Some risk processes have been described that integrate the two.

Transdiagnostic Risk Factors

Negative affect and depression

The experience of subjective distress or negative affect predicts disordered eating, including binge eating, in youth (Leon, Fulkerson, Perry, Keel, & Klump, 1999; Pearson, Zapolski, & Smith, 2014b). It also predicts binge eating in laboratory studies (Agras & Telch, 1998) and relates to eating disorder attitudes and cognitions (Casper, Hedeker, & McClough, 1992; Klump, McGue, & Iacono, 2002; Leon et al., 1993). Moreover, women with BN are more likely to binge eat on days during which they experience negative emotions, and the binge episodes follow periods of escalating negative emotion (Haedt-Matt & Keel, 2011; Smyth, Wonderlich, Heron, Sliwinski, Crosby, et al., 2007). Similar findings have accrued for depression as a predictor of disordered eating (Pearson et al., 2014b; Stice, Burton, & Shaw, 2004; Tanofsky-Kraff et al., 2011). In addition, 50% of women with BN are actively depressed when they are first seen for an evaluation and 80% have a lifetime history of depression (Herzog, Keller, Zacks, Yeh, & Lavori, 1992). A large portion of the overlap between negative affect and binge eating/emotional eating is explained by genetic factors (Racine, Keel, Burt, Sisk, Neale, Boker, & Klump, 2013).

Related to the impact of negative affect on disordered eating is the construct of emotional eating, which can be defined as “eating in response to a range of negative emotions such as anxiety, depression, anger, and loneliness, to cope with negative affect” (Tanofsky-Kraff, Theim, Yanovski, Bassett, Burns, Ranzenhofer, Glasofer, et al., 2007). Those who engage in emotional eating are more likely to experience a loss of control, or feel as if they cannot stop, when they eat during the experience of negative emotions (Tanofsky-Kraff et al., 2007). This finding indicates the important role emotions play with respect to disordered eating. Emotional eating is present in children and adolescents (Shapiro, Woolson, Hamer, Kalarchian, Marcus, & Bulik, 2007; van Strien, 1996; van Strien, Engels, Van Leeuwe, & Snoek, 2005) and evidence suggests a relationship between emotional eating and episodes of binge eating in adults (Eldredge & Agras, 1996; Greeno, Wing, & Shiffman, 2000; Kendary, Arnow, & Agras, 1996; Masheb & Grilo, 2006; Tanofsky, Wilfley, Spurrell, Welch, & Brownell, 1997; Telch & Agras, 1996).

Two sources of negative affect have been identified as particularly important for women with BN: Problems in interpersonal relationships and negative self-evaluation. Concerning relationships, whether assessed by observation or self-report, adults and adolescents with BN report more conflicted, disengaged, and non-nurturant relationships (Humphrey, 1986, 1987, 1989; Kendler, Maclean, & Neale, 1991; Strober & Humphrey, 1987; Wonderlich, 1992). Concerning negative self-evaluation, women with bulimic symptoms display elevated levels of perfectionism, particularly maladaptive forms of perfectionism which include doubts about actions and concerns about mistakes (Bardone-Cone, Joiner, Crosby, Crow, Klein, et al., 2008). In addition, women with BN tend to perceive a discrepancy between their actual self-concept (how they see themselves) and a comparative ideal standard that they apply to themselves or believe others apply to them (Strauman & Glenberg, 1994; Strauman, Vookles, Berenstein, Chaiken, & Higgins, 1991; Wonderlich, Engel, Peterson, Robinson, Crosby et al., 2008). This discrepancy between the actual self and the ideal self is experienced as a deficit. For women with BN, self-discrepancies are often in the domains of body dissatisfaction and body size (Strauman et al., 1991; Strauman & Glenberg, 1994; Wonderlich et al., 2008).

Clearly, negative affect confers transdiagnostic risk. Constructs such as negative affect and subjective distress are included as diagnostic criteria in more disorders than are any other constructs in DSM-5 (American Psychiatric Association, 2013). Depression is highly comorbid with many other disorders, and through the shared experience of negative affect, a variety of psychiatric disorders may form a broader class of emotional disorders, including other forms of emotional dysfunction such as anxiety as well as many forms of personality dysfunction (Goldberg, Krueger, Andrews, & Hobbs, 2009).

Emotion regulation and distress tolerance

Women with BN appear to have limited capacity to tolerate emotional distress (Anestis, Selby, Crosby, Wonderlich, Engel, et al., 2010). Furthermore, they may have broader deficits in emotion regulation, particularly in terms of identifying, clarifying, experiencing, and accepting various emotional states (Lavender, Startup, Naumann, Samarawickrema, Dejong et al., 2012). Deficits in emotion regulation and distress tolerance have clear transdiagnostic implications. For example, they have been implicated in borderline personality disorder (Gratz, Rosenthal, Tull, Lejuez, & Gunderson, 2010), non-suicidal self-injury (Gratz, 2007), and substance use (Wills, Pokhrel, Morehouse, & Fenster, 2011).

Impulsivity and negative urgency

We define impulsivity with respect to behavior: Impulsivity refers to rash behavior that is characterized by a focus on meeting one’s immediate need, or acting on an immediate urge, without due consideration of the possible negative consequences of the act with respect to one’s long-term interests, goals, or health. There are many different personality traits that relate to impulsive behavior, and a number of different systems for identifying relevant traits (Barrett, 1993; Buss & Plomin, 1975; Cyders & Smith, 2007; 2008a; Depue & Collins, 1999; Dickman, 1990; Gray & McNaughton, 2000; Whiteside & Lynam, 2001; Wills et al., 2011). There are also ongoing advances in further understanding relevant impulsivity traits, including on the neurobiological level (Fineberg, Potenza, Chamberlain, Berlin, Menzies, et al., 2010; Robbins, Gillan, Smith, de Wit, & Ersche, 2012).

One particular impulsivity-related trait that has proven important in the prediction of bulimic behaviors is the trait of negative urgency, which refers to the tendency to act rashly or impulsively when distressed (Cyders & Smith, 2007, 2008a; Whiteside & Lynam, 2001). Negative urgency distinguishes individuals with BN from controls (Cyders, Smith, Spillane, Fischer, Annus, & Peterson, 2007) and predicts the onset of binge eating and purging among both early adolescents (Pearson, Combs, Zapolski, & Smith, 2012a) and college women (Anestis, Selby, & Joiner, 2007; Fischer, Peterson, & McCarthy, 2013).

When negative urgency is examined with other impulsivity-related traits, including those that reference sensation seeking and failure to plan, it is the only trait that consistently predicts binge eating (Anestis, Smith, Fink, & Joiner, 2009; Claes, Candereycken, & Vertommen, 2006; Fischer & Smith, 2008) and purging (Combs, Pearson, & Smith, 2011; Pearson, Combs, & Smith, 2010). In a meta-analysis, negative urgency had a weighted effect size for concurrent prediction of bulimic symptoms of r = .38 (effect sizes for other impulsivity-related traits ranged from .08 to .16: Fischer, Smith, & Cyders, 2008). A large portion of the overlap between the impulsivity-related trait of negative urgency and binge eating/emotional eating is explained by genetic factors (Racine et al., 2013). Negative urgency and negative affect have both shared and independent genetic overlap with binge eating and emotional eating (Racine et al., 2013).

Given negative urgency’s relationship with negative affect, it may also be functionally related to low distress tolerance, although this possibility has not yet been fully explored. Engaging in rash acts when distressed is thought to facilitate the avoidance of distress (Cyders & Smith, 2009). Consistent with this possibility, negative urgency correlates highly with low distress tolerance (Combs & Smith, 2012).

Negative urgency appears to confer transdiagnostic risk. In prospective research, the trait predicts subsequent increases in heavy drinking, smoking, risky sex, drug use, gambling, and other pathological behaviors (Anestis et al., 2007; Cyders, Flory, Rainer, & Smith, 2009; Cyders & Smith, 2008b; Doran, Khoddam, Sanders, Schweizer, Trim, & Myers, 2012; Fischer et al., 2013; Settles, Cyders, & Smith, 2010; Zapolski, Cyders, & Smith, 2009) in addition to binge eating and purging.

Eating Disorder-Specific Risk Factors

Thin ideal internalization and body dissatisfaction

Through advertisements and many other forms of social communication, women are consistently exposed to messages that thinness is good and if one is thin, one will be happy and successful (Garner, Garfinkel, Schwartz, & Thompson, 1980; Rubinstein & Caballero, 2000). Not surprisingly, women who internalize this thin ideal tend to experience body dissatisfaction, due to the unrealistic nature of the ideal. Internalization of the thin ideal and body dissatisfaction are understood to increase the following: (a) risk for excessive dieting (in an effort to meet the ideal); (b) binge eating precipitated in part by dietary restriction, as described further below, and (c) negative affect due to body dissatisfaction and failure to meet one’s ideal (Catarin & Thompson, 1994; Field, Camargo, Taylor, Berkey, Frazier, et al., 1999; Field, Camargo, Taylor, Berkey, Roberts, et al., 2001; Pyle, Mitchell, & Eckert, 1981; Stice, 2001; Stice & Bearman, 2001; Stice, Mazotti, Krebs, & Martin, 1998; Stice, Presnell, & Spangler, 2002; Striegel-Moore, Silbertstein, & Rodin, 1986; Wertheim, Koerner & Paxton, 2001). Body dissatisfaction predicts bulimic symptom onset (Field et al., 1999; Killen, Taylor, Hammer, Litt, Wilson, et al., 1993; Stice & Agras, 1998) and increases in bulimic pathology (Cooley & Toray, 2001; Stice, 2001).

Dieting

Women often diet or deprive themselves of calories in an effort to lose weight and change their body shape/size. Repeated efforts to restrict food intake and failures of those efforts predict increases in negative affect (Stice & Bearman, 2001; Stice, Mazotti, Weibel, & Agras, 2000) and bulimic symptoms (Field et al., 1999; Killen et al., 1994; Killen, Taylor, Hayward, Haydel, Wilson et al., 1996; Stice, 2001; Stice & Agras, 1998).

Expectancies for reinforcement from eating and from thinness

Expectancies are understood to represent summaries of individuals’ learning histories, based on a multitude of direct and vicarious learning experiences that individuals undergo (for basic science expectancy accounts, see Bolles, 1972; Goldman, Brown, Christiansen, & Smith, 1991; Rotter, 1975; Tolman, 1932). One tends to engage in behaviors from which one expects rewards and avoid behaviors for which one expects punishment. To the extent that one comes to associate eating with more powerful reinforcers than do others, one will then hold unusually strong expectancies for reinforcement from eating. One therefore will pursue food with greater vigor. To the extent that one forms stronger expectations that thinness brings a range of reinforcers, one will pursue thinness more vigorously. It is noteworthy that these expectancies result in completely opposing behavioral tendencies.

Strikingly, women with BN simultaneously endorse both the expectancy that eating helps manage negative affect and the expectancy that thinness leads to overgeneralized life improvement more strongly than comparison women. Specifically, they endorse the eating expectancy at higher levels than do women with AN, normal controls, and psychiatric controls. They endorse the thinness expectancy at higher levels than do normal and psychiatric controls, and at similar levels as do women with AN (Bruce, Mansour, & Steiger, 2009; Hohlstein, Smith, & Atlas, 1998). Their strong expectations for reward from both eating and thinness is consistent with their engagement in extreme eating, dieting, and purging behaviors.

In prospective research, endorsement of the expectancy that eating helps manage negative affect predicts binge eating frequency as well as binge eating onset in samples of adolescent girls (Pearson et al., 2012a; Smith et al., 2007; Stice & Whitenton, 2002) and college women (Fischer et al., 2013). Endorsement of the expectancy that thinness leads to overgeneralized life improvement predicts increases in both purging and binge eating in middle school girls (Smith et al., 2007). In the laboratory setting, reduction in this thinness expectancy leads to a reduction in body dissatisfaction and cognitive symptoms of disordered eating (Annus, Smith, & Masters, 2008).

We believe that identification of expectancies for reinforcement from both eating and thinness offers a useful way to integrate the classic risk factors for BN described above. Because human behavior is heavily influenced by the pursuit of reward (Herrnstein, 1990), we view the learned anticipation or expectation of reward as an important mechanism by which exposure to environmental events leads to behavior. From this perspective, endorsement of the expectancy that eating helps manage negative affect can be understood to reflect a summary of learning experiences relating eating to reward in the form of negative reinforcement that contributes to emotional eating, and hence to loss of control eating. Endorsement of the thin-ideal reflects the learned expectancy that thinness will bring reward in the form of positive reinforcement. The reason to endorse the thin-ideal is that one expects thinness to bring positive benefits not otherwise available. Because the societal ideal of thinness is not realistic, women who have come to associate such thinness with reward tend ultimately to become dissatisfied with their bodies; body dissatisfaction can thus be viewed as a “down-stream” consequence of strong endorsement of the expectation that thinness will bring important rewards.

Existing Integrative Theories

Theories of BN have existed in the literature for many years and have provided fruitful explanations for bulimic behaviors. These theories were developed before the identification of some of the risk factors reviewed above, including negative urgency and eating and thinness expectancies, and also before several basic science advances (e.g., self-control theory) that we will consider later in this paper.

Restraint Model

One of the most well-known theories of risk and maintenance for BN is the restraint model (Polivy & Herman, 1985; Striegel-Moore et al., 1986). This theory proposes that, due to strong societal pressures for a thin physique, women are at risk to begin dieting and restricting food intake. Once they begin restricting their food intake, their bodies feel deprived. They begin craving food, thus increasing the likelihood that they will binge eat. Of course, a binge eating episode is not consistent with their goal of losing weight, so after a binge episode they purge in an effort to prevent the weight gain they anticipate as a result of the binge eating episode. The women then begin restricting again in an effort to “recover” from the binge, thereby restarting the cycle. Thus, the restraint model describes a process that includes explanations for both the initiation of bulimic behaviors and their maintenance.

Important to the restraint model are the risk factors of internalization of the thin ideal and dieting. Because the “thin ideal” environment appears to be ubiquitous, at least for European American women (Polivy & Herman, 1985; Striegel-Moore et al., 1986), an important challenge for the restraint model has been to understand which women may develop BN in response to the environment, given that the vast majority do not. The model does not specify individual difference or psychosocial learning processes that differentiate among high and low-risk women.

Dual-Pathway Model of Bulimic Pathology

The dual pathway model (Stice, 2001; Stice & Agras, 1998) holds that internalization of the thin ideal is associated with heightened levels of body dissatisfaction. It also adds that the pressure to be thin, delivered through relationships and cultural norms, further contributes to body dissatisfaction. Body dissatisfaction is thought to foster dieting and negative affect, which together increase the likelihood of bulimic symptomatology. The model further posits that individuals may engage in bulimic behavior either because of extreme dieting or chronic negative affect or, importantly, a combination of the two. Either of the two pathways is sufficient to promote the onset of bulimic behavior.

Emotion Regulation and Escape Models

The emotion regulation model is quite different from the restraint and dual pathway models. In this model, the trigger for bulimic events is understood to be negative affect. Women who experience high levels of negative emotions frequently are at risk to binge eat in an effort to alleviate the negative emotion (e.g., Agras & Telch, 1998; Gratz & Roemer, 2004; Wonderlich, Peterson, Mitchell, Crow, Smith et al., 2014). The sources of negative affect may or may not involve eating disorder-related factors, such as body dissatisfaction or food deprivation. Whatever the source of the negative affect, binge eating episodes and purging are thought to provide reward in the form of negative reinforcement: Negative emotions decrease after the binge or purge is complete. Ecological momentary assessment, or EMA, research documents increasing negative affect leading up to binge-purge episodes in women with BN, followed by decreasing negative affect after the episodes (Smyth et al., 2007).

Escape theory (Heatherton & Baumeister, 1991) is a similar model which posits that individuals who binge eat are prone to perceive inadequacies in themselves which generate negative emotion. Similar to the emotion regulation model, bulimic behavior evolves as a response to negative affect and helps to alleviate negative affect. However, escape theory differs from emotion regulation theories in positing that negative emotions decrease during the binge eating episode as opposed to after the binge eating episode. That is, during the binge episode, women are distracted, experience decreased self-awareness and cognitive narrowing to immediate stimuli, and hence disinhibition (Heatherton & Baumeister, 1991). The binge episode is negatively reinforced through the distraction from distress; however, as self-awareness returns following the binge eating episode, negative emotions are thought to increase (Heatherton & Baumeister, 1991). The fundamental process described by escape theory need not be specific to BN: Many other behaviors that are considered disinhibited or impulsive, such as substance abuse, suicidal behavior, and self-harm can provide negative reinforcement when distressed (Heatherton & Baumeister, 1991).

The emotion regulation and escape models offer two things not provided by the restraint or dual-pathway models. The first is the focus on the transdiagnostic risk factor of negative affect. (The dual-pathway model does not identify trait-oriented negative emotion as a risk factor; rather, the negative emotion emerges secondary to body dissatisfaction: Stice, 2001.) The second is an account of one way in which individual differences in personality influence risk for BN. The possibility that bulimic behaviors serve as a potent affect-based reinforcer for women who tend to be higher in negative affectivity is an appealing one, as researchers seek to explain how BN is maintained despite the many harms it brings.

Along with the strengths of the emotion regulation and escape models are two potentially important shortcomings. First, these models do not specify the factors that result in the choice of bulimic behaviors to regulate emotion or escape from distress. When faced with significant negative affect, there are many possible behavioral options to cope, some more adaptive and some less adaptive. Why would someone engage in binge eating and purging in the first place? Related to this, the models do not systematically address characteristics of the environment that contribute to BN, nor do they include specific psychosocial learning mechanisms (other than negative reinforcement) that increase risk for BN.

Second, these theories appear more useful for understanding the maintenance of bulimic behaviors over time than their initiation. Once a person has experienced distress relief from bingeing and purging, the negative reinforcement process is a compelling explanation for why those behaviors would be repeated in the future and, over time, become more and more frequent. In contrast to the restraint and dual-pathway models, which offer an explanation for initiation, the emotion regulation and escape models offer stronger explanations for maintenance of bulimic behaviors over time.

A New, Integrative Theory for BN: From Impulsive Action to Compulsive Behavior

In brief, our theory incorporates a variety of risk factors and processes which are related to the fundamental proposition that emotion-based processes are inherently involved in the promotion of bulimic behavior. Women’s risk for BN is heightened by the operation of a psychobiological system involving emotion-based impulsivity. One personality-based expression of this system is the trait of negative urgency, which predicts multiple forms of psychopathology and which has been described in terms of its psychobiological components. In part as a result of elevations in negative urgency and in part due to psychosocial influences, some women form unusually strong expectancies for reinforcement both from eating and from thinness. Elevations in both types of expectancies increase the likelihood of bulimic behavior. We understand these processes to operate on a trait level. Additionally, consistent with other theories, we hold that state level processes will influence bulimic behavior through self-control depletion due to efforts to manage negative affect and suppress the biological drive to eat. Self-control depletion results in a momentary experience of emotion-based impulsivity in the form of binge eating and purging. A crucial aspect of our theory is the idea that while BN behavior begins as an emotion related impulsive action, it ultimately acquires compulsive features in which the behavior functions to avoid distressing events and negative emotions in spite of the individual’s awareness of the significant harm that is brought by the behavior (see Koob, 2009; Koob and Le Moal, 1997 for definitions of compulsivity).

Neurobiological Underpinnings for Emotion-Based Impulsive Action

We begin by briefly considering the neurobiological underpinnings for impulsive behavior and self-control depletion. There is considerable evidence to support the view that emotional processing prepares the body for action (Frijda, 1986; Lang, 1993; Maxwell & Davidson, 2007; Saami, Mumme, & Campos, 1998); emotional experiences lead to activation of the motor cortex (Hajcak, Molnar, George, Bolger, Koola, et al., 2007; Morgenson, Jones, & Yim, 1980). Although emotion’s facilitation of action is fundamentally adaptive in that emotions signify needs that can be met by actions, it is often necessary to exercise self-control over the urge to act immediately when experiencing intense emotion. Whereas intense negative emotions can lead to a focus on actions that provide immediate relief from the precipitating distress, it is often necessary to inhibit immediate, affect-driven responses and formulate responses that are consistent with one’s ongoing goals, interests, or well-being. Thus, adaptive responses to intense distress typically involve problem-solving approaches concerning the source of the distress together with an ongoing awareness of and connection to one’s longer-term interests and goals (Davidson, 2003).

Although problem-solving responses to distress are typically adaptive, intense distress can undermine adaptive responding. It can undermine rational decision making (Bechara, 2004, 2005; Dolan, 2007; Dreisbach, 2006), in part by interfering with one’s orientation toward the pursuit of long-term goals and by increasing distractibility (Dreisbach & Goschke, 2004). In recent years, research conducted in several different domains has helped clarify possible underlying brain mechanisms for this process. This research has identified a relevant functional brain system and has highlighted the roles of key neurotransmitters within that brain system.

Neural Systems

There appears to be a functional brain system involved in the processing of emotion-laden experiences and preparing for action involving interconnections between the amygdala and the orbitofrontal cortex (OFC) and its medial sector (the ventromedial prefrontal cortex, or VMPFC) (Barbas, 2007; Bechara, Tranel, & Damasio, 2000; Ghashghaei & Barbas, 2002; LeDoux, 2000; M. D. Lewis & Todd, 2007). The amygdala appears to be involved in the experience of emotionally salient stimuli, perhaps particularly negative affect; it is activated in response to such stimuli (Davidson, 2003). As we describe below, the OFC and VMPFC appear to be involved in the modulation of emotion-based reactivity (Davidson, 2003). Both areas receive direct projections from the sensory areas (Barbas, 2007), from one another, and from the anterior cingulate cortex (ACC: Devinsky, Morrel, & Vogt, 1995).

When the amygdala is activated in response to emotionally arousing stimuli, its projections to the OFC/VMPFC operate. Sometimes called “bottom-up” processing (M.D. Lewis & Todd, 2007), these projections appear to have the effect of alerting the OFC/VMPFC to attend to the emotionally important stimuli. At the same time, projections from the amygdala to the striatum, the nucleus accumbens, and the ventral tegmental area enhance the activation of both limbic and cortical structures, thereby further preparing for action in response to the emotional experience (Cardinal, Parkinson, Hall, & Everitt, 2002). This sequence of processing from the amygdala to higher level cortical areas helps orient one to what is important and also helps one prepare possible behavioral responses.

When the OFC or VMPFC receives input from the amygdala, sensory areas, and the ACC, its projections back down to the amygdala and other areas are activated, with the apparent result of regulating the amygdala and the brain stem (Bechara, 2005; Ghashghaei & Barbas, 2002; Hariri, Drabant, & Weinberger, 2006; M. D. Lewis & Todd, 2007); this is often referred to as “top-down” processing. The ACC appears to be involved in monitoring conflicts between intended and actual responses (Inzlicht & Gutsell, 2007). Thus, the PFC structures provide information about the anticipated consequences of possible actions, with a bias toward long-term, goal-directed behavior. The OFC/VMPFC can override the emotional response of the amygdala (M. D. Lewis & Todd, 2007), with two effects: (a) shortening the time course of the experience of negative affect and attention to stressful stimuli and (b) reducing the likelihood of emotion-driven rash responses to the emotion and thus increasing the likelihood of responses that are consistent with one’s ongoing interests and goals (Davidson, 1998, 2003).

In a well-functioning system, individuals recognize emotionally salient stimuli and attend to them. Based on cortical input, they choose behaviors designed to meet the needs of the underlying emotion but in a way that does not harm their long-term interests or goals. The process of choosing actions based on one’s ongoing interests and goals, rather than choosing actions designed only to respond to the immediate affect requires what is known as effortful control (Rothbart, Ahadi, & Evans, 2000), which is thought to involve the maintenance of an emotional connection to one’s long-term goals and interests (Davidson, 2003).

It is important to note that our theory posits that both situational and trait-like factors can lead to fewer cautioning signals from the OFC/VMPFC to the amygdala (Donegan, Sanislow, Blumberg, Fulbright, Lacadie, et al., 2003; Silbersweig, Clarkin, Goldstein, Kernberg, Tuescher, et al., 2007). One result is higher levels of emotion-driven action that are less informed by one’s long-term interests (Cyders & Smith, 2008a); failures of inhibition can occur (Carver, Johnson, & Joormann, 2008). Concerning situational factors, there is considerable evidence that self-control over amygdala-driven affectivity and rash action is limited and becomes depleted after initial self-control exertions (Baumeister & Heatherton, 1996; Muraven, Tice, & Baumesiter, 1998; Vohs & Heatherton, 2000). On a neural level, it appears to be the case that the ACC supports a conflict-monitoring system, which monitors ongoing behavior and is sensitive to discrepancies between intentions and actual responses (Inzlicht & Gutsell, 2007). Through the ongoing exertion of self-control (e.g. restricted eating, emotional avoidance), such resources are believed to be depleted and their neural systems tend to be less responsive to a mismatch between their actions and their goals, as reflected in reductions in EEG responses that originate from the dorsal ACC that reflect monitoring of such mismatches (Inzlicht & Gutsell, 2007). Weakened identification of such mismatches provides less information to the PFC and results in less PFC-based modulation of amydala activity. This set of circumstances increases risk for momentary failures of self-control within a person.

Alternatively, such deficits in self-control can be attributed to trait-like or stable variations in the system. On the most basic level, individuals with damage to the OFC/VMPFC display an impaired ability to effectively anticipate the potential risks of their actions (Bechara, Damasio, Tranel, & Damasio, 2007). OFC/VMPFC-damaged individuals are particularly sensitive to immediate reinforcement and punishment and oblivious to future consequences of their actions (Bechara, Damasio, Damasio, & Anderson, 1994). In such individuals, failures of self-control are relatively stable and trait-like, but behaviorally they resemble the state-based depletion scenario described previously. One important contributor to such OFC/VMPFC activity is the presence and activity of key neurotransmitters, which can contribute to malfunction in this system on both the state and trait levels, thereby setting the stage for impulsive action.

Neuromodulators

Of course both bottom-up and top-town interactions between the amygdala and OFC appear to occur through neuromodulators like serotonin (5HT) and dopamine (DA; M. D. Lewis & Todd, 2007). Experimental and correlational designs have demonstrated that low levels of 5HT are associated with greater rates of sensation seeking and reckless behaviors together with greater levels of both negative and positive affect (Cools, Blackwell, Clark, Menzies, Cox, et al., 2005; Depue & Collins, 1999; Frankle, Lombardo, New, Goodman, Talbot, et al., 2005; Krakowski, 2003; Morgan, Impallomeni, Pironi, & Rogers, 2006; Spoont, 1992; Winstanley, Dalley, Theobald, & Robbins, 2004a; Winstanley, Eagle, & Robbins, 2006; Winstanley, Theobald, Dalley, Glennon, & Robbins, 2004b; Zald & Depue, 2001). Considering the brain system of interest, low levels of 5HT result in lower levels of OFC/VMPFC functioning and therefore fewer top-down transmissions from the OFC/VMPFC back to the amygdala in the presence of strong emotion. This results in lesser inhibition of the amygdala, a stronger and more lasting emotional response, and reduced ability to consider the possible long-term consequences of actions. One result is lower self-control capacity, or an increased disposition toward disinhibited or rash action (Cyders & Smith, 2008a).

For some receptor subtypes and in some brain areas, 5HT modulates DA activity (Morelli, Moore, Rebello, Gray, Steele, et al., 2011; Spoont, 1992). In those cases, when low levels of 5HT are present, DA levels are less inhibited, thereby increasing levels of DA in the brain. Thus, areas of one’s brain can have low levels of 5HT and high levels of DA. Elevated levels of DA are associated with a tendency to act (Depue, 1995; Depue & Collins, 1999; Spoont, 1992; Zald & Depue, 2001) and increasing reward-seeking and risk-taking behaviors (Spear, 2000). High levels of DA activity in the amygdala-OFC/VMPFC circuit are associated with high rates of rash or reckless acts (Floresco & Tse, 2007). One result of this combination is an overall reduction in OFC/VMPFC modulation of amygdala-driven responses, or a reduction in effortful controland an increased likelihood of engaging in rash, ill-considered acts (Cyders & Smith, 2008a; Depue, 1995).

Although this process is a general one, there is evidence that it applies specifically to BN. Women with BN tend to have low levels of the major metabolite of 5HT (Kaye, Gwirtsman, George, & Ebert, 1991; Steiger & Bruce, 2007). Impaired serotonergic responsiveness is often cited during the acute illness state of BN (e.g., Brewerton, Lydiard, Laraia, Shook, & Ballenger, 1992; Goldbloom, Garfinkle, Katz, & Brown, 1990; Kaye, Greeno, Moss, Fernstrom, Fernstrom, et al., 1998) and is also associated with BN symptom severity (Jimerson, Lesem, Kaye, & Brewerton, 1992). Moreover, higher levels of impulsivity in women with BN are associated with low levels of 5HT (Steiger, Israel, Gauvin, Ng Ying Kim, & Young, 2003). The rash BN behaviors of binge eating and purging therefore appear associated with variations in brain systems that are associated with reduced effortful control.

Two Risk Pathways for Impulsive BN Behavior

We identify two risk pathways that can lead to initial, impulsive engagement in bulimic behaviors. The first is state-based and focuses on risk as a result of momentary behaviors and emotional experiences. The second is trait-based and focuses on risk as a function of individual differences among women. We propose that both risk pathways involve both transdiagnostic and eating disorder-specific risk processes.

Situational, State-Based Pathway: Self-Control Depletion and Eating Disorder Risk

Several characteristics of high-risk women make them vulnerable to engagement in bulimic behaviors as a result of momentary behaviors and emotional experiences. First, these women are likely to experience negative affect, often due to difficulties in their interpersonal relationships or pathologic self-evaluation and self-regulation (Wonderlich et al., 2008). In particular, negative self-evaluations in relation to the thin ideal, and hence body dissatisfaction, can result in a momentary experience of negative affect (Stice, 2001).

Second, due to individual differences in psychosocial learning histories, including parental and peer modeling, women vary in their development of expectancies that eating helps alleviate negative affect and that thinness provides overall life improvement. Parental modeling in the form of modeling snacking, modeling binge eating, and modeling mood improvement from eating is associated with stronger expectancies that eating helps alleviate negative affect in daughters (MacBrayer, Smith, McCarthy, Demos, & Simmons, 2001). Parental expressions of concern over a daughter’s weight and reports of having been teased about one’s weight as a child are associated with stronger expectations that thinness leads to overgeneralized life improvement as well (MacBrayer et al., 2001). Both types of expectancies are formed prior to the emergence of binge eating and purging (Pearson et al., 2012a; Smith et al., 2007). High-risk women tend simultaneously to expect reinforcement from eating and from thinness.

When high-risk women are also dieting in pursuit of the thin ideal, they are facing numerous demands on their self-control resources. Restricting food intake, which works against a basic biological drive for food, requires self-control. The self-control demands associated with food restriction are all the greater when women are distressed and hold the expectancy that eating will help alleviate their distress. Also, the experience of negative emotions tends to deplete self-control, because self-control efforts are required to inhibit immediate, non-adaptive responses to the negative emotion (Tice, Bratsalavsky, & Baumeister, 2001).

The basic science literature has provided a useful perspective on factors that contribute to the depleted ability to exercise self-control for people in general. Repeated exercise of self-control leads to a reduced capacity to continue self-control activities in the moment (Baumeister, Heatherton, & Tice, 1994; Muraven & Baumeister, 2000). Thus, the self-control demands associated with momentary experiences of negative affect and food depletion, in the context of high-risk expectancies, increases the likelihood of subsequent self-control failures, and thus of binge eating and subsequent purging.

There may be an additional contributor to momentary depleted ability to control food intake that has not yet been directly investigated. It may be the case that restricting food intake makes learned expectancies for reinforcement from eating more salient, with the result that expectancies play a stronger role in influencing momentary behavior. Perhaps dietary restriction can be understood to activate existing eating expectancies, such as the expectancy that eating helps manage distress, which heightens the risk for loss of control eating when self-control mechanisms fail. Related to this, it is also possible that the relief associated with eating when one feels deprived strengthens the expectancy. If expectancies are shaped by psychosocial learning, it is possible that the efforts to restrict food intake, and consequent binge eating, may have important influences on subsequent expectancy change.

Following the binge eating episode, these women are also at increased risk to purge. Their binge eating behavior undermines their goal of thinness that is reflected in their strong endorsement of the thin ideal. Thus, their expectancies for reinforcement from thinness, fears about weight gain, high levels of emotional distress, and depleted self-control, along with their physical discomfort, put them at increased risk to take the rash, impulsive step of purging the food they have consumed. They are less likely to engage in more measured, effortful control-driven processes to change their eating and dieting habits to reduce risk in the long run. Consequently, following heavy self-control demands (e.g., attempting to restrict eating while experiencing a negative mood and having negative self-thoughts), women may experience momentary depletions in self-control, thereby increasing risk for engaging in a binge eating and purging episode.

Trait-Based Pathway: Stable Dispositions and Eating Disorder Risk

Just as momentary self-control depletion processes may lead to dyscontrolled acts at the state level, there is a personality disposition that disposes individuals toward engagement in emotion-driven rash or dyscontrolled acts at the trait level. Negative urgency is defined as a personality trait that reflects individual differences in the propensity to engage in rash, impulsive acts when distressed (Cyders & Smith, 2008a; Pearson et al., 2012a; Whiteside & Lynam 2001).

As mentioned previously, there is considerable evidence that negative urgency operates as a transdiagnostic risk factor, in that it prospectively predicts increases in, or the onset of, numerous maladaptive behaviors (Anestis et al., 2007; Cyders et al., 2009; Cyders & Smith, 2008b; Doran et al., 2012; Fischer et al., 2013; Pearson et al., 2012a; Settles et al., 2010; Zapolski et al., 2009). Negative urgency is consistently associated with deficits in self-control, reflected in various forms of rash, impulsive action. Interestingly, consistent with the Cyders and Smith (2008a) urgency theory hypothesis, the 5HT gene polymorphism associated with reduced cortical modulation of emotion-driven activity is associated with elevations in negative urgency (Carver, Johnson, Joorman, Kim, & Nam, 2011).^{1, 2}

Thus, an important part of the trait-based risk pathway toward engagement in the impulsive acts of binge eating and purging reflects ongoing, stable deficits in self-control that can be measured as negative urgency. The consistent findings of both cross-sectional associations and longitudinal prediction from negative urgency to the subsequent onset of, or increases in, binge eating and purging in both children and college students support this claim (Anestis et al., 2007; Combs et al., 2011; Fischer et al., 2013; Pearson et al., 2012a).

Negative urgency and the Acquired Preparedness model of risk

Furthermore, negative urgency may also influence expectancies that BN behavior may be rewarding. Both cross-sectional and prospective research document the additive predictive effects of negative urgency, expectancies for reinforcement from thinness, and expectancies for reinforcement from eating in the prediction of binge eating and purging behaviors (Combs et al., 2011; Pearson et al., 2012a). We next describe a process by which negative urgency and eating disorder-specific risk processes transact to increase risk for bulimic behaviors.

Above, we described evidence that environmental events, such as parental modeling and peer influences, likely shape the formation of eating and thinness expectancies. In addition, there is theory and evidence indicating that personality traits in general, and negative urgency in particular, also help shape the learning process. It has long been recognized that personality interacts with environmental events to influence subsequent behavior (Caspi, 1993; Caspi & Roberts, 2001). For example, individuals with different personalities may react differently to the same stimulus. A person high in neuroticism is more likely to perceive an event as stressful than is an emotionally stable individual.

An extension of this idea involves the demonstration that individuals with different traits actually have a different learning outcome from precisely the same learning experience; in its general form, this idea traces back as least as far back as Rotter (1954). That is, individuals who vary in their trait make-up but who experience the exact same outcome in a task form different expectancies about the likely reinforcement from the task (Smith, Williams, Cyders, & Kelley, 2006). Traits help shape the learning process; individuals with different traits are disposed to learn different things. Because one’s learning is a proximal influence on one’s behavior, it follows that the influence of personality traits on behavior is mediated by psychosocial learning. Thus, personality may influence the formation of expectancies for reinforcement from eating and from thinness.

This risk process has been termed acquired preparedness (AP), to denote the theory that individuals differentially acquire high risk expectancies (i.e., preparedness) as a function of high-risk personality traits (Smith & Anderson, 2001; Smith et al., 2006). Applied to BN, the AP model of risk is as follows (Combs & Smith, 2009). Individuals high in negative urgency are disposed to act to alleviate distress; often, their actions are rash or impulsive (Settles, Fischer, Cyders, Combs, Gunn, & Smith, 2012). When individuals high in this trait are exposed to learning experiences (whether direct or modeled) that involve eating when distressed, they are biased to form expectancies associating the behavior of food consumption with the negative reinforcement of distress relief (Fischer et al., 2008). This expectancy, as noted above, is understood to be a proximal risk factor that increases the likelihood of binge eating behavior (Smith et al., 2007a). Thus, over time, negative urgency predicts stronger endorsement of the expectancy that eating will help alleviate distress, and the expectancy endorsement, in turn, predicts subsequent binge eating behavior. This process has been shown in longitudinal studies of both young girls (Pearson et al., 2012a), and college women (Fischer et al., 2013).

The Combined Influence of State and Trait Factors on Bulimic Behavior

To understand how bulimic behaviors occur for a given person, it is necessary to understand how elevations in trait risk factors can result in engagement in bulimic behaviors at any one moment. The model we have described leads directly to hypotheses integrating individual difference and momentary risk processes. We offer both a mediation hypothesis and a moderation hypothesis to describe how state and trait factors transact to increase risk. Neither of these hypotheses has yet been tested.

Concerning mediation, momentary deficits in self-control are posited to mediate negative urgency’s effect on bulimic behavior. To the degree that women high in negative urgency respond to distress with rash actions more often than do others, they have fewer learning trials in which to develop self-control capacities. As a result, their self-control capacities may be weaker than is true for others. Thus, those women who are high in negative urgency and experience distress, perhaps triggered by the environment (e.g., relationship stress) or eating disorder behavior (e.g., food restriction) simply have a harder time engaging in self-regulation; they experience self-control depletion sooner. Endorsing the expectancy that eating behavior will alleviate their distress, they are at increased risk for binge eating compared to others. Holding the expectancy that thinness will lead to overgeneralized life improvement, they are at increased risk to then purge. In this way, binge eating and purging may serve as an attempt to regulate affect that was driven by momentary self-control depletion, which occurred in part due to elevations in the trait of negative urgency. That is, self-control depletion mediates the influence of the trait of negative urgency on BN behavior.

Concerning moderation, we propose the possibility of an interaction between trait-based negative urgency and state-based self-control depletions. Women can experience self-control depletion for many reasons, including negative affect related to relationship distress or food restriction. We suggest that elevated levels of negative urgency heighten the risk for binge eating and purging among women experiencing such momentary depletion of self-control. If women high in negative urgency who are experiencing momentary self-control depletion expect eating to alleviate negative affect, they are more likely to binge eat. If they expect life improvement from thinness, they are more likely also to purge. By contrast, this model would suggest that even women with elevated levels of negative urgency are less likely to act rashly when their self-control resources have not been depleted. Thus, in the absence of negative mood or extensive dieting attempts, negative urgency alone is less likely to lead to bulimic behaviors. This set of hypotheses follows plausibly from the existing negative urgency and self-control literatures. Again, though, they have yet to be tested (although see Wonderlich, Crosby, Engel, Mitchell, Smyth, & Miltenberger, 2007 and Engel, Boseck, Crosby, Wonderlich, Mitchell and colleagues, 2007 for evidence relating trait impulsivity to momentary negative moods).

Reinforcement from Binge Eating and Purging Episodes in the Early Phase of the Disorder

To this point, we have attempted to integrate the concepts of negative urgency, self-control depletion and expectancy in our theory of bulimic behavior. However, to further account for the emergence and maintenance of BN, it is important to understand why early, impulsive bulimic behaviors are repeated in the future. If they are repeated by the same person over time, it is very likely that they provide some form of reinforcement. We next consider the nature of reinforcement from bulimic behaviors.

The timing of the presumed negative reinforcement for bulimic behaviors is a matter on which eating disorder researchers disagree. One of the most innovative and useful ways to study behavior and its immediate effects is through ecological momentary assessment (EMA), which involves multiple reports over the day of mood, behavior, and other variables. In a large sample of women with BN, Smyth et al. (2007) used this method to assess binge eating episodes and found that affective states significantly worsened (negative affect slowly increased and positive affect slowly decreased) before a binge and then there was a recovery with negative affect decreasing and positive affect increasing following a binge. The rate of affective recovery post-binge was significantly more rapid than the rate of affective decay pre-binge (Smyth et al., 2007); thus, the authors suggested that the affective change is unlikely to be the result of transient situational or external influences. On days when women with BN did not binge, the pattern of increasing and then decreasing negative mood (or decreasing and then increasing positive mood) was not observed (Smyth et al., 2007). This finding is consistent with other literature (Cattanach, Malley, & Rodin, 1988; Crowther & Sherwood, 1997; Johnson & Larson, 1982; Kjelsas, Borsting, & Guddle, 2004; Lingswiler, Crowther, & Stephens, 1989; Schlundt, Johnson, & Jarrell, 1985; Waters, Hill, & Waller, 2001; Yacono Freeman & Gil, 2004) and is thought to provide direct evidence that binge eating episodes provide negative reinforcement (as well as positive reinforcement).

A different approach from the EMA approach of examining trajectories of affective change prior to and following bulimic episodes is to use a single point approach and measure affect at a moment prior to an episode and then again immediately following an episode. A meta-analysis using this approach found that negative affect was higher at the post-episode measurement than the pre-episode measurement (Haedt-Matt & Keel, 2011). This meta-analytic finding suggests the possibility that women with BN are in fact more distressed just after they binge than they are preceding a binge. Thus, the single point findings differ from the EMA affective trajectory findings. As a result, questions remain about the process by which early bulimic behaviors are reinforced.

One possible explanation for the discrepancy is that pre-episode measurements of negative affect may occur well before the episode, whereas post-episode measurements often occur within minutes of the episode. Recently, Berg, Cao, Crosby, Engel, Peterson, et al. (2014) examined three EMA studies and found that the most recent pre-episode affective rating occurred approximately two hours before the episode, and the post-episode rating occurred on average 2–9 minutes after the episode. Curves fitting the trajectories of affect in these 3 EMA studies produced the trajectories of affective change pattern described above, but when the single point estimates were derived for negative affect two hours prior to the behavior and 2–9 minutes after the behavior, they were consistent with the Haedt-Matt and Keel (2011) findings. Thus, it is possible that single point estimates may reflect an increase in negative affect from pre- to post-bulimic behavior but still coincide with a curve depicting an overall decrease in negative affect due to differences in the timing of the pre- and post- assessments relative to the bulimic behavior. Further research, perhaps using methods that do not rely on self-report, is necessary to elucidate the reinforcement process.

With these competing perspectives in mind, we offer the following account of reinforcement associated with bulimic behaviors in the early phase of the disorder. We posit that the experience of binge eating provides immediate reinforcement that occurs during, and possibly immediately after, the binge episode. Similarly, purging provides prompt negative reinforcement as well.

Reinforcement through Distraction

We believe the first experience of reinforcement from BN behavior does not occur after the binge is complete; rather, it is much more immediate and occurs in the early stages of the binge, or even as one prepares to binge. During the binge episode, one is eating favored foods (typically, foods high in fat and sugar), and there is a loss of control associated with the consumption (APA, 2013). These behaviors are likely to distract one from sources of personal distress, and are thus negatively reinforced (similar to Heatherton & Baumeister, 1991). Indeed, an experimental intervention in which distraction from distress was not permitted during binge episodes led to significant reductions in binge eating among women with BN or Binge Eating Disorder (Bosch, Miltenberger, Gross, Knudson, & Breitwieser, 2008; Giddings & Miltenberger, 2010). Because binge eating episodes often involve consumption of favored foods (often, foods one has deprived oneself of during a period of food restriction), they are positively reinforcing as well. Just prior to a binge eating episode, one is engaged in the tasks of choosing and selecting favored foods. Thus, cognitive resources are shifted toward favored foods and the binge eating episode, and away from sources of distress.

One recent pilot study used an indirect method of assessing reinforcement in order not to undo whatever distraction is occurring during the planning stage (Pearson, Anderson, Powell, Chester, Terry, & Smith, 2012b). While in a functional magnetic resonance imagining (fMRI) scanner, women with BN were placed in a negative mood and brain activity was compared when they were planning a binge by selecting pictures of food on which to binge, versus planning a future apartment by selecting pictures of furniture. The authors found a large drop in amygdala activation when women with BN were planning a binge, compared to almost no drop when they selected furniture. This pattern of amygdala activity is associated with a drop in negative affect (Urry, van Reekum, Johnstone, Kalin, Thurow, et al., 2006). They also found increased activation in the caudate, insula, putamen, and rostral anterior cingulate cortex, which may indicate both positive reinforcement through reward and negative reinforcement through the reduction of distress when planning a binge versus engaging in a neutral task while distressed (Pearson et al., 2012b).

The clear limitations of this pilot work highlight the need for further empirical inquiry into the negative reinforcement by distraction hypothesis. Most importantly, the sample size was small (n = 9). As a result, although the effect sizes were all large (d > 1.0), comparisons were not statistically significant. The second limitation is that the study involved women with established BN, so the results cannot speak directly to reinforcement during the early stages of bulimic behaviors. The third limitation is the use of an indirect method to assess negative and positive reinforcement. Thus, there is not as yet sufficient empirical evidence to confirm the immediate reinforcement through distraction hypothesis that we have presented.

Negative Reinforcement of Purging Behavior

We hold that purging behavior provides very prompt negative reinforcement for women following a binge eating episode. Binge eating behavior typically represents a failure of efforts to pursue the thin ideal, resulting in cognitive, emotional, and physical distress. When a girl purges, she believes she has removed much of the food consumed, thus reducing her caloric intake from the binge. This step might well provide some immediate relief from the physical discomfort following the binge and the distress regarding concerns about weight gain, even though purging does not rid the body of all binge related calories (Kaye, Weltzin, Hsu, McConaha, & Bolton, 1993). This contention is consistent with the meta-analysis by Haedt-Matt and Keel (2011), who found a reduction in negative affect following purging events.

To summarize, we hold that binge eating and purging behaviors provide immediate negative reinforcement for women who are in the early stages of engaging in bulimic behaviors, and that binge eating behavior provides positive reinforcement as well. Although we believe there is good support for this position, we also believe there is a clear need for further inquiry into the nature and timing of reinforcement from early bulimic behaviors.

From the Impulsive to the Compulsive: The Maintenance of BN

Just as binge eating and purging behavior is described as impulsive, it is also described as compulsive: Women with the diagnosis of BN repeatedly engage in the binge-purge cycle despite knowing and experiencing the harms associated with the behavior. Continued engagement in a behavior despite knowing that doing so is harmful is referred to as compulsive (Skodol & Oldham, 1996). One important challenge for any model of BN is to account for both its initially impulsive nature and the eventual compulsive quality that characterizes more enduring cases of the disorder. To present our account, we will first review established neurobiological models for the compulsive use of drugs, because there is good reason to believe similar processes may operate for BN. We then present an integration of the impulsive and compulsive with respect to BN.

Neurobiological Processes that Appear to underlie Compulsive Behaviors: A Drug Abuse Model

Each time an individual uses a drug that has positive, reinforcing properties, there is an increase in dopaminergic activity, which has been shown to be associated with subjective perceptions of reward (Volkow, Wang, Fischman, Foltin, Fowler, et al., 1997). These feelings of pleasure reinforce the drug use so the individual comes to choose it as a way to increase positive affect and decrease negative affect. However, the neuroplasticity associated with the drug use causes an elevation in the brain reward threshold, thereby creating a greater and greater elevation in “baseline” thresholds of reward (Koob & Volkow, 2010; Solomon, 1980). Because the reward function fails to return within normal homeostatic range, the individual needs more of the drug to achieve the same effect of immediate reward and pleasure. In addition, when the reward system is chronically activated by drug use, the brain adapts by decreasing the number of dopamine D₂ receptors, which results in a decreased sensitivity of the reward circuits to stimulation by natural reinforcers, like receiving praise, positive social interactions, or going for a walk (Martin-Solch, Magyar, Kunig, Missimer, Schultz, & Leenders 2001; Volkow & Fowler, 2000).

As a result of these processes, after extensive, ongoing drug use, one experiences less frequent positive affect and comes to find little pleasure in everyday activities and instead craves the drug and focuses on seeking the drug. Furthermore, the chronic activation of the reward system also activates an opponent process (Solomon, 1980): An anti-reward circuit that has the opposing actions of limiting the drug’s reward function (Koob, 2009; Solomon, 1980).

The combination of the loss of function of natural reward systems and the recruitment of the brain stress or anti-reward systems creates a situation of chronic loss of pleasure and ongoing negative affect; these states can only be alleviated by ever-increasing amounts of the drug on an escalating frequency of occasions. Thus, use of the drug becomes a powerful source of negative reinforcement. Individuals choose to continue using the drug to achieve the negative reinforcement, despite knowing that the use has adverse consequences: They use the drug compulsively (Koob & Bloom, 1988; Koob & Le Moal, 1997; Koob & Volkow, 2010).

Application of the Drug Abuse Model to Food and BN

There are neurobiological parallels between drug use and food consumption. Both tend to result in dopaminergic activity within the reward centers of the brain. In fact, when BN individuals view pictures of palatable foods, they show similar brain activations in the insula and caudate as do individuals addicted to substance when they view pictures of choice drugs (Pelchat, Johnson, Chan, Valdez, & Ragland, 2004). Exposure to sweet tasting foods results in an overflow of DA in the nucleus accumbens, which appears to initiate the reward system response and drive to consume (Adam & Epel, 2007; Cota, Barrera, & Seeley, 2006; Hajnal, Smith, & Norgren, 2004). When individuals do actually binge eat on sweet, rewarding foods, DA is repeatedly released in response (Avena, Rada, & Hoebel, 2006).

Women with BN tend to have a lower number of DA receptors in the brain’s reward systems, and there is a significant negative association between the frequencies of binge eating and purging and the striatal DA response (Broft, Shingleton, Kaufman, Liu, Kumar, et al., 2012). Such evidence suggests that because of this decreased DA response, more DA needs to be released to achieve an adequate reward value (Volkow, Wang, Fowler, Logan, Jayne, et al., 2002; Volkow, Wang, Maynard, Jayne, Fowler, et al., 2003). Thus, similar to what is observed in individuals with substance use disorders (Volkow et al., 2002, 2003), individuals with BN tend to have lower baseline levels of responsiveness to DA, which is associated with reduced baseline levels of pleasure or reward, resulting in less positive affect and also requiring larger upsurges in DA to experience reward states.

Given these similarities, we anticipate that a variant of the drug model applies for bulimic behavior as well. When women engage in bulimic behavior during the early, impulsive stage, they experience a rush of DA, which is perceived as rewarding. This pleasant feeling reinforces the bulimic behavior so it becomes more likely in the future. Each time the individual chooses to engage in BN behavior when distressed, more and more DA is needed to reach that same reward threshold.

Despite having an already low number of DA receptors, the chronic elevation of DA activated by bulimic behavior requires a down regulation of the receptors, resulting in decreased sensitivity of the reward circuits. As the positive reinforcement associated with bulimic behavior decreases through down regulation processes, the behavior continues to provide negative reinforcement through distraction and reduction of negative affect. The behavior is now compelled and serves a function as an affect regulation strategy to moderate emotional states.

Bulimic Behaviors Serve an Avoidance Function for Women with BN

We believe there is another important aspect of the nature of bulimic behavior as it progresses to the compulsive. Over time, with repeated conditioning trials, the binge-purge event comes to serve an anticipatory avoidance function, rather than just the escape function of distracting one from distress or the function of providing positive reward. A process of this kind appears to operate for repeated rewards in general. Initially, behaviors might provide immediate positive reinforcement (perhaps related to the release of DA) and negative reinforcement (such as escape or distraction from distress). Over time, the same behavior can serve to avoid anticipated distress, which can come from recruited stress systems as described in opponent-process theory (Solomon, 1980) or from the sources of distress from which one escaped when initially engaging in the behavior (such as body dissatisfaction or relationship difficulty).

This idea has been applied fruitfully to drug abuse. Baker, Piper, McCarthy, Majeskie, and Fiore (2004) argue that, through repeated cycles of drug use, addicted individuals learn to identify cues of emerging negative affect before they experience the negative affect itself. These individuals are then motivated to pursue and use their drug of addiction to avoid the emergence of painful negative affect. For them, the urge to use drugs precedes the experience of subjective distress, and if the urge is gratified, the distress is avoided.

In a similar way, we believe that women with BN come to associate the urge to engage in bulimic behaviors with distress cues more and more automatically and earlier in the distress cycle. Over time, they become aware of feeling the urge to binge eat, or to binge eat and purge, rather than experiencing the distress associated with negative events.

In a recent, unpublished study involving intensive interviews with women with BN, Pearson and Smith (2013) found that these women frequently described a growing urge to engage in bulimic behaviors that continued to build over the course of the day until they engaged in the behaviors. This finding is consistent with other BN literature regarding urges (Steiger, Lehoux, Gauvin, 1999; Verstuyf, Vansteenkiste, Soenens, Boone, & Mouratidis, 2013). Often, women with BN start to plan the episode when it is not yet possible to engage in it (consistent with Abraham & Beaumont, 1982). For example, during the work day, they plan what food to buy and when they will binge eat, and they feel an increasing urge to engage in bulimic behavior that has not been gratified. When the authors asked them specific questions about distressing events or negative affect, they described such events, indicating the presence of distress, but they reported that their focus remained on the urge.

The Binge-Purge Cycle for Women who have an Entrenched, Compulsive Bulimic Pattern of Behavior

In the hours leading up to a bulimic event, women with relatively enduring BN are able to identify an increasing urge to engage in bulimic behavior (Pearson & Smith, 2013). Their focus on this urge provides for them a focus on a need that can be met, rather than on the distressing life events and negative affect that are also present. In this way, the urge serves an avoidance function; it distracts from the present negative affect and distressing situation. Women with BN can plan the bulimic episode, which creates a subjective sense of stability and control that seemingly keeps emotional distress at bay. In this sense, experiencing an urge to engage in bulimic behaviors provides the negative reinforcement of reduced focus on distress.

Of course, negative affect is still present, but the function of focusing on the urge avoids the subjective experience of the negative emotion. Thus, if women with BN are asked how they feel and are required to focus their attention back on their feelings, they may report the presence of and the gradual increase in negative affect leading up to a bulimic episode (see: Haedt-Matt & Keel, 2011 and Smyth et al., 2007).

Following the bulimic episode, individuals with BN have gratified the urge, but they also feel considerable physical discomfort from the binge (or binge-purge), as well as guilt and shame for having gratified an urge in a maladaptive manner that works against their goals. Over the ensuing hours, they start to feel better physically and the improved affective state associated with gratifying an urge becomes more salient. Therefore, it may be the case that bulimic behavior becomes a conditioned response to an urge to binge eat and purge, which was once a conditioned response to negative affect and which now may facilitate avoidance of the momentary experience of negative affect. This process provides women with BN with a sense of controllability and stability as they can gratify the urge by engaging in the symptomatic behavior.

At some point, after repeated cycles of emotional distress – urge – bulimic behavior, the behavior is compelled. These women are able to recognize that binge eating and purging is not an adaptive response and will bring harms, but they continue to engage in the behaviors because the fundamental value of the behavior is so powerful in terms of distress avoidance and affect regulation. It is also important to appreciate that, to the degree that this cycle provides consistent reinforcement initially, over time it is likely to take on the characteristics of a habit that is relatively immune to the intermittent absence of reinforcement (Walsh, 2013). For these reasons, over time, binge eating and purging becomes compelled.

Further Issues

Considerations for Ethnic Minority Groups

As we noted above, the bulk of the research that forms the basis of our theory was conducted on samples of women who primarily came from European American origins. A great deal remains to be learned about the similarities and differences in eating disordered behavior and risk processes across ethnic groups (Jacobi et al., 2004; Striegel-Moore & Bulik, 2007). Some research suggests that eating disorder symptoms may vary by ethnicity. Hispanic American adolescents appear to have higher rates of BN than girls of other ethnicities (Swanson, Crow, Le Grange, Swendsen, & Meikangas, 2011). In one study, purging behaviors in the absence of binge eating were more common in European American women than African American women, whereas binge eating in the absence of purging was more common among African American women (Striegel-Moore et al., 2005). However, other research suggests that recurrent binge eating prevalence does not vary by ethnicity (e.g., Reagan & Hersch, 2005; Smith, Marcus, Lewis, Fitzgibbon, & Schreiner, 1998; Striegel-Moore, Wilfley, Pike, Dohm, & Fairburn, 2000).

Given the dearth of research in this area, it is not surprising that there are few conclusions that can confidently be drawn about differences or similarities in the process by which eating disorders develop across ethnicities. Many studies have examined the risk factors in sociocultural models of eating disorders across cultures and found that: (a) perceived pressures to be thin and body dissatisfaction are lower among Hispanic, African American, and Asian women than European American women (Flynn & Fitzgibbon, 1996; Kemper, Sargent, Drane, Valois, & Hussey, 1994; McKnight Risk Factor Study, 2000; Neff, Sargent, McKeown, Jackson, & Valois, 1997; Striegel-Moore, Wilfley, Caldwell, Needham, & Brownell, 1996; Winkelby, Gardner, & Taylor, 1996); (b) dieting is lower among African American and Asian American individuals than European Americans (French, Story, Neumark-Sztainer, Downes, Resnick, et al. 1997; Langer, Warheit, & Zimmerman, 1991); and (c) internalization of the thin ideal was less pronounced in African American and Hispanic females than in Asian American or European American females (Shaw, Ramirez, Trost, Randall, & Stice, 2004).

The preceding findings refer to mean levels of risk factors. A different, and perhaps more important, question concerns cultural differences in the relationships among risk factors or between risk factors and symptomatology. There is some evidence that risk factors drawn from the sociocultural framework differ in magnitude across ethnic groups. For example, one study found that dietary restraint was predicted by different risk factors in different ethnic groups and that the CBT restraint model, incorporating constructs from the sociocultural model, was only useful for European American girls (White & Grilo, 2005). Other researchers found that relationships between interoceptive awareness and internalization of the thin ideal, and between internalization of the thin ideal and body dissatisfaction were positive for European American, Mexican American, and Spanish women, but that the relationships were significantly stronger for European American women than the other ethnic groups (Warren, Gleaves, Cepeda-Benito, Fernandez, & Rodriguez-Ruiz, 2005). It would of course not be surprising if there were ethnic differences in relationships among constructs understood to be sociocultural in nature.

For the current theory, there is evidence to suggest that the negative urgency process described here may be similar for at least two ethnic groups, European Americans and African Americans (Pearson et al., 2012a). In addition, the factor structures of measures reflecting expectancies for reinforcement from eating, reinforcement from dieting and thinness, and bulimic symptomatology have been found to be invariant across those two groups (Atlas, Smith, Hohlstein, McCarthy, & Kroll, 2002). African American women had lower mean scores on both the risk factors and the measure of bulimic symptomatology, but correlations between risk factors and symptoms did not differ across race. Socioeconomic status did not account for these results; both groups were college students from the same university. Bardone-Cone and Boyd (2007) also studied college women and found lower mean levels of both risk and symptom measures for African American women than for European American women. Similarly, African American children endorsed the risk factors in the AP model (negative urgency and expectancies for reinforcement from eating) for binge eating at a lower rate than did European American children, but again the magnitude of the associations between those risk factors and binge eating did not differ across the groups: The AP risk model was invariant across the two groups (Pearson et al., 2012a). These findings suggest that the risk and maintenance model presented here may be similar for African American and European American females, with the total level of dysfunction perhaps being lower for African American women. Other findings (e.g., White & Grilo, 2005) suggest that some aspects of the risk process, like restraint and thin-ideal internationalization, may in fact be different across ethnic groups. More research is needed in this area in order to better understand the developmental process across ethnicities.

Clinical Implications

To the extent that our theory is supported in future work, it will be important to develop and test new treatments for BN that incorporate the risk and maintenance factors presented in this model, including an emphasis on transdiagnostic risk associated with negative urgency, together with an emphasis on eating disorder-specific risk factors, such as expectancies for reinforcement from eating and from thinness. Based in part on portions of the theory we have presented, Wonderlich and colleagues (2014) developed Integrative Cognitive-Affective Therapy for Bulimia Nervosa (ICAT-BN), an integrative approach to treating BN that emphasizes the momentary relationship between situational cues, emotions, and bulimic behaviors, particularly in terms of the rewarding value of the behaviors. Specifically, ICAT addresses emotion-based eating, negative self-evaluation, self-criticism, interpersonal problems, and regular eating (i.e., normalized eating patterns similar to individuals without disordered eating), all of which we have deemed important in the present model, especially those related to the state based loss of control pathway presented.

ICAT-BN (Wonderlich et al., 2014) also includes interventions that target negative urgency, with an emphasis on: (a) enhancing emotional awareness and emotion regulation through describing and labeling of emotions; (b) identifying cues for emotional experiences; (c) learning how to tolerate distress in the moment; (d) increasing positive emotions; and (e) learning alternative means of coping that do not harm oneself. In part, emotion-focused treatments for BN (e.g., dialectical behavior therapy (DBT: Linehan, 1993); ICAT) can help women with BN better understand their emotions, what prompted them, their functions, and how they are experienced. As a result, they can better identify and understand their urge to binge eat and purge, thereby allowing for behavioral modifications and alternative means of coping. Furthermore, increasing positive and pleasurable events in their lives will help to increase reward and positive emotions, which would help promote healthy DA release and reduce vulnerability to negative emotions in the long run. Possibly, a focus on negative urgency in treatment is likely to have ancillary benefits for women with BN, as women learn alternative strategies for responding when experiencing intense negative emotion.

In the context of this focus on negative urgency and affect regulation, treatments may be further enhanced by considering more fully how learned expectancies about eating and thinness may influence an individuals’ anticipation that eating alleviates distress and that thinness leads to overgeneralized life improvements. A direct focus on the anticipated reinforcement associated with both eating and thinness, and development of alternative ways to achieve such reinforcement, may provide a useful adjunct to existing treatments.

Conclusion

We have presented a risk and maintenance model for BN that builds on existing models and, through the integration of literature advances in several fields, offers a new theoretical account. By relying on recent advances in psycho-neurobiology, self-control depletion theory, and personality theory, we specify several transdiagnostic processes that contribute to risk for both onset and maintenance of the disorder. By relying on existing research on eating disorder-specific risk processes, as well as advances in emotion regulation and eating disorder expectancy research, we have described risk processes that involve transactions between global and eating disorder-specific factors. We specify both a state-based risk process and a trait-based risk process, and we offer hypotheses concerning mediation and moderation relationships between the two. We argue that the bulimic behaviors of binge eating and purging are initially impulsive in nature, and that over the course of progression to diagnosable BN, those behaviors come to be compelled responses to distress or the risk of distress. We hope that this model proves generative of both new empirical investigation and improved theory development, thus leading to advances in understanding this serious psychological disorder.