Loss of control eating (LOC; i.e., eating accompanied by a sense of not being able to control what or how much one is consuming) is a transdiagnostic eating disorder (ED) symptom associated with considerable psychological distress (1), functional impairment (2), and reduced quality of life (3). LOC also plays a key role in perpetuating cycles of disordered eating. The transdiagnostic cognitive-behavioral model of EDs suggests that concerns about weight gain following LOC can lead to increased restriction, compensatory behaviors, and poor body image, which can in turn lead to more LOC in a maladaptive cycle (4). LOC and related disorders [e.g., bulimia nervosa (BN), binge-eating disorder (BED)] also represent a substantial public health concern. Approximately 10% of women, and a smaller but significant number of men, experience clinically significant binge eating (5). While there are empirically supported treatments for LOC, many who receive treatment continue to experience symptoms (6).

Though current DSM-5 diagnostic criteria focus on objective binge eating (OBEs; i.e., LOC over an objectively large amount of food), other forms of LOC and dysregulated eating that do not meet the formal definition of an “objective binge” in DSM-5 [e.g., subjective binge eating (SBEs), or LOC over relatively small amounts of food] may be even more prevalent (7) and associated with equivalent distress and comorbidity (8, 9). People with OBEs and people with SBEs are indistinguishable on numerous relevant variables, including dietary restraint, cognitive ED symptoms (e.g., body image concerns), disinhibition, general psychopathology, impairment, health service utilization, and endorsement of having an eating problem (9, 10). In contrast, overeating without LOC is associated with significantly less distress and impairment than OBEs and SBEs (11). Correspondingly, ICD-11 criteria treat OBEs and SBEs as equivalent, with either behavior sufficient to meet the binge eating criterion for BN and BED (12). Factor analytic studies also indicate that disordered eating is dimensional rather than categorical (e.g., OBEs and SBEs may represent different points on a dysregulated eating continuum) (13). In recognition of evidence regarding the dimensional nature of LOC, and evolving diagnostic standards that treat OBEs and SBEs as diagnostically equivalent, this review will examine a broad construct of LOC that includes OBEs, SBEs, and emotional eating (i.e., eating in response to negative emotions). Though emotional eating is not explicitly defined as involving loss of control, it shows robust empirical associations with other forms of LOC (14, 15), and similar patterns of change in response to negative affect (16) and biological fluctuations [e.g., ovarian hormone levels (17)], suggesting it is a closely related behavior. By incorporating multiple forms of LOC, this review will be poised to directly examine whether reproducible differences in the factors maintaining LOC over different amounts of food have been found.

Understanding the etiology and maintaining factors underlying LOC is crucial for advancing research and treatment. Several prominent etiologic theories highlight the role of negative affect (NA; i.e., negative emotions, including guilt, anger, and sadness) and emotion regulation difficulties in promoting LOC. These theories are consistent with evidence of elevated LOC among individuals with internalizing disorders (e.g., anxiety, depression) (18), and increased rates of internalizing disorders among people with EDs characterized by LOC (19). In other words, LOC may be one manifestation of more general negative emotionality and underlying difficulties with managing negative emotions among people with LOC-related disorders. This conceptualization has led to adaptation of treatments developed for other disorders characterized by affect dysregulation, such as dialectical behavior therapy (DBT), to BN and BED (20). Interestingly, treatments such as DBT that focus on more general emotion dysregulation appear to decrease comorbid psychopathology and other maladaptive behaviors (such as self-injury) in addition to LOC (21), suggesting that addressing affect dysregulation may be crucial to improving both core ED symptoms and accompanying concerns that contribute to distress/impairment.

Several affect regulation models have been developed, each of which proposes a somewhat different mechanism underlying the association between NA and LOC (see Figure 1). In its most straightforward iteration, the standard affect regulation model of LOC posits that LOC is triggered by an increase in NA, and is negatively reinforced by a decrease in NA following LOC (22). Operationally, this theory predicts that NA should be greatest just before LOC, and decrease following LOC. Related theories have proposed slightly different functions of LOC and trajectories of NA surrounding this behavior. For example, escape theory suggests that LOC serves to distract an individual not only from general NA, but specifically from negative awareness or attributions about the self (23). Like the standard affect regulation model, escape theory predicts that NA is elevated prior to LOC and triggers the onset of this behavior. However, escape theory differs from the standard affect regulation model in predicting that NA only decreases while the eating episode is ongoing, and increases again following LOC once self-awareness is reestablished.

Three additional theories diverge from the standard affect regulation model in hypothesizing that NA may trigger LOC, but LOC does not necessarily downregulate NA in turn. Restraint theory proposes that NA temporarily disrupts the cognitive control required to maintain food restriction, leading to dysregulated eating (24). In this case, NA could potentially continue to increase following LOC. Alternatively, trade-off theory posits that LOC may replace a highly aversive negative emotion (e.g., anger) with one that is more easily tolerated (e.g., sadness, guilt) (25). The overall intensity of NA may therefore remain unchanged following LOC, but the specific type of emotion experienced may shift. Finally, expectancy theory hypothesizes that beliefs regarding the affect regulatory properties of eating may drive LOC to a greater extent than the actual consequences of this behavior. Specifically, expectancy theory proposes that the expectation that one will feel better after eating increases the likelihood of eating to relieve NA, even if one's mood does not actually improve (26).

While all affect regulation theories predict that elevated NA may trigger LOC, they diverge in their predictions regarding trajectories of NA during and after LOC. The affective consequences of LOC are difficult to measure with the granularity necessary to distinguish between affect regulation theories using questionnaires administered once or at a limited number of timepoints, and may be distorted by the artificial and unfamiliar laboratory context in experimental research. Instead, changes in affect during and after LOC are optimally examined using experience sampling designs, which ask participants about their emotions and eating behavior in the context of their everyday lives. During experience sampling, questionnaires are administered once per day for several days (in daily diary studies) or multiple times per day [in ecological momentary assessment (EMA) studies], providing information about changes in emotions over relatively short timespans. Experience sampling designs have been used to study dynamic affective processes across EDs and other forms of psychopathology (27), and provide unique data regarding trajectories of affect in the minutes or hours before, during, and after LOC. Experience sampling designs also allow for examination of individual differences in momentary affect-LOC associations. For example, it is possible to test the hypothesis put forth by expectancy theory that individuals who more strongly believe that eating will reduce NA will in fact be more likely to eat in response to negative emotions.

This paper will comprehensively review evidence from experience sampling studies (including both daily diary and EMA designs) that have examined NA-LOC relationships to determine which affect regulation theory or theories are best supported by current research. The review will address three primary questions. First, what do the data say regarding changes in NA before and after LOC in naturalistic settings (i.e., does NA tend to increase, decrease, or remain stable in the minutes and hours before and after LOC)? Second, are certain negative emotions (e.g., guilt and anger) particularly likely to precede LOC, and do these differ from the negative emotions that are most likely to follow LOC (as predicted by trade-off theory)? Third, are there individual differences in momentary NA-LOC associations, and, if so, do they align with existing affect regulation theories (e.g., expectancy theory)?

Studies published prior to January 1, 2021 were considered for inclusion. Relevant studies were identified by searching PubMed and PsycINFO using the terms (“binge eating” or “emotional eating” or “loss of control eating” or “dysregulated eating”) and (“ecological momentary assessment” or “experience sampling” or “daily diary”). Additional studies were identified through manual search of the reference lists of prior meta-analyses [e.g., (22)]. Studies were included if they used an EMA or daily diary design to examine associations between NA and OBEs, SBEs, or emotional eating in adults 18 or older. Children and adolescents were not included in the present review due to evidence of developmental differences in the etiology (28) and phenomenology (29) of LOC and related disorders in youth.

Three primary experience sampling designs have been used to examine changes in affect surrounding LOC in daily life. Daily diary designs, in which participants report on their affect and eating behavior once at the end of the day, are the least intensive. While daily diary designs represent the lowest burden to participants, they also provide less fine-grained data than more intensive designs. Conversely, daily diary designs are better suited to detecting longer-term associations between NA and LOC (e.g., over a period of days or weeks) than briefer, more intensive studies. Intermediate in intensity are EMA designs that collect information on a participant's affect at several semi-randomly distributed times throughout the day, signaled by an alert on a participant's cellphone or a similar device (i.e., signal-contingent ratings). Some EMA studies also ask participants to immediately report when they experience LOC (i.e., event-contingent ratings). Similar to daily diary designs, these kinds of EMA studies can be used to examine whether NA is elevated near LOC episodes, but provide greater temporal precision (i.e., a window of a few hours rather than an entire day). The most intensive experience sampling designs ask participants to report on NA at multiple timepoints during LOC itself. These are the only designs that provide information about a participant's emotions while LOC is in progress. However, this method risks altering the natural consequences of LOC by directing participants to focus on their emotions during the behavior.

Results from each of these study designs will be discussed, then synthesized with respect to their implications for affect regulation models of LOC. Study details (e.g., sample size and demographics, operationalizations of NA and LOC) and key findings for each study are summarized in Appendix A in Supplementary Material, along with an assessment of study quality (described in Appendix B in Supplementary Material).

Experience sampling research to date presents evidence of complex associations between NA and LOC in daily life that do not completely adhere to the predictions of the standard affect regulation model of LOC. Across study designs, elevated or increasing NA is associated with increased risk for LOC on a given day (30, 31, 33) and in subsequent hours [e.g., (43–45, 48, 50, 52–54, 56), perhaps particularly for individuals with frequent and recurrent LOC (109) and OBEs (37, 79, 92). However, the preponderance of evidence suggests that LOC may not be an effective strategy for decreasing NA, either immediately after the behavior or across longer timespans. Designs that have measured NA immediately before and after LOC eating have consistently found that NA is greater after LOC than before LOC (76–79, 94). While NA begins to decrease in the hours following LOC (e.g., (64, 80–85, 87)], research has often found that it remains elevated over baseline (64, 65, 90), and that LOC on 1 day tends to predict higher NA on the next (30, 31). Purging after LOC may decrease NA, but only to levels observed before LOC (102, 104). Overall, results from experience sampling studies challenge the view that LOC is maintained primarily through its ability to reduce negative emotions.

Even if LOC does not improve mood, people who generally expect to feel better after eating may be particularly vulnerable to LOC when they experience even small increases in NA (82). Greater expectations that eating will help one feel better in a given moment also strengthen the association between NA and LOC a short time later (48, 53). Expectations that eating will help one feel better may be particularly likely to lead to LOC when combined with other state risk factors for LOC, such as food restriction (53). Research findings to date are therefore consistent with expectancy theory, and highlight the potential role of cognitive factors (e.g., beliefs about the affective consequences of eating) in the maintenance of LOC. Interestingly, the two-way interaction between dietary restraint and NA was not significant in Pearson et al. (53), suggesting that restraint may only amplify the association between NA and subsequent LOC when a person also believes that eating will help them feel better. However, additional studies are needed to more fully examine whether restraint might weaken cognitive control in the presence of NA and contribute to LOC through the mechanisms proposed by restraint theory.

If people often feel worse after LOC, why does this behavior persist? Escape theory and trade-off theory offer two potential explanations by proposing that NA and self-critical thoughts may decrease during LOC (23), or that LOC may replace one negative emotion with another that is less aversive (25). Findings from experience sampling research have been mixed for both these theories. While some studies have found that NA temporarily decreases during LOC (100, 101), an equal number have found that NA remains constant or increases across LOC (102, 103). One study suggests that people with greater overall dissociative tendencies experience more rapid decreases in NA after LOC (87). However, Mason et al. (87) did not examine emotions or dissociation during LOC itself, and it therefore remains unclear whether a dissociative experience during LOC may temporarily reduce NA, as predicted by escape theory. Rather than replacing more aversive negative emotions with less unpleasant ones, LOC seems most likely to decrease low arousal negative emotions (e.g., boredom) and increase high arousal emotions that may be more distressing (e.g., guilt, anger at self) (101, 104). Neither theory therefore provides an unambiguous explanation for the persistence of LOC, at least based on current research.

One possible alternative explanation is that people with LOC may mistakenly attribute the gradual decrease in NA following LOC to the act of eating itself. People with EDs are more likely than people without disordered eating to endorse maladaptive beliefs about emotions, such as the idea that emotions will last forever or become overwhelming (111). In other words, individuals with LOC may fear (at least in the moment, when negative emotions are intense) that they will never feel better if they do not engage in LOC, or at least do something to regulate their affect. Interestingly, similar maladaptive beliefs about emotions are common in anxiety disorders, such as the belief that emotions will become so strong that they cannot be tolerated in panic disorder (112). Individuals with these beliefs may fail to recognize that negative emotions will eventually decrease on their own, and could potentially misattribute the natural decrease in NA over time following a distressing event to LOC. To test this theory, it would first be necessary to definitively establish that the decrease in NA following LOC is in fact no faster than the decrease in NA we would expect in the hours after a peak in negative emotions (e.g., after an unpleasant event) not accompanied by LOC. It would also be informative to conduct additional research on the beliefs about emotions endorsed by people directly before and after LOC (e.g., the extent to which individuals believe they will feel better on their own over time if they do nothing, the extent to which they believe LOC helped them feel better). If the evidence suggests that LOC does not hasten a decline in NA, but people with LOC strongly believe that they must act to feel better, individuals with LOC may benefit from psychoeducation and experiential exercises demonstrating that NA will frequently decrease over time on its own, even if they do nothing at all.

A complementary potential explanation is that, unless explicitly asked, people with LOC may not pay close attention to how their emotions change pre- to post-eating. Alexithymia, or difficulty identifying and communicating feelings, is common among people with OBEs (113) and other forms of LOC (114). If a person is not closely attuned to how their emotions change when they experience LOC, it may be easier to maintain the belief that LOC decreases NA, even when this is not the case. This may be particularly true in a cultural context where social messages about the soothing effects of eating (e.g., “comfort food”) are ubiquitous (115), and the “default belief” may be that eating helps reduce NA. In this case, mindfulness of one's own emotional experience, and mindful attention to how that experience may differ from common cultural beliefs, could potentially help decrease a person's tendency to engage in LOC to relieve NA.

It is also possible that the effects of LOC on NA may change over time. Specifically, LOC may effectively decrease NA early in the course of an illness such as BN or BED, but begin to increase NA as LOC becomes more frequent or persistent, and its associated undesirable consequences become clearer. LOC may then persist as a habitual response to NA that becomes difficult to change even though it is no longer reinforcing. Additional research on individuals who have recently begun experiencing LOC (perhaps recruited from high-risk populations or large, population-based samples of adolescents or young adults) vs. those who have experienced this behavior for months or years could illuminate differences in the emotional impact of LOC across time.

Affect regulation theories of LOC have traditionally focused on changes in NA, with less attention to the effects of LOC on positive affect (PA; e.g., joy, excitement, or contentment). Note that PA is not merely the absence of negative emotions—correlations between well-validated measures of NA and PA are small [e.g., −0.12 to −0.23 on the Positive and Negative Affect Schedule (105)]. It is therefore possible that someone could feel worse (i.e., more NA) and better (i.e., more PA) after LOC. This is particularly true given that consumption of palatable foods is intrinsically pleasurable, especially after a period of food restriction (116). Most studies that have examined NA and PA separately have found that PA is lower or unchanged immediately after LOC (65, 70, 74), arguing against maintenance of LOC via positive reinforcement [but see (104), who found an increase in happiness/relief following purging]. However, relatively few studies have investigated PA, and additional research on changes in PA surrounding LOC in daily life is warranted. In particular, it is possible that increases in PA following LOC may be particularly likely if LOC was proceeded by dietary restriction. This possibility could be investigated by examining prior restriction as a moderator of increases in PA following LOC in future research.

Three additional limitations of existing experience sampling research on associations between NA and LOC should be acknowledged. A first key limitation is inherent to the methodology—namely, that intensive monitoring of NA surrounding LOC may itself alter NA-LOC associations. If one function of LOC is to escape awareness of NA, drawing attention to emotions may negate this effect. This is particularly true for more intensive EMA designs that sample affect at multiple timepoints during LOC itself. One potential solution is to incorporate affect measures into future EMA research that do not require active reporting by participants. While physiological sensations do not correlate perfectly with verbal reports of emotion (117), non-invasive monitoring of heart rate or skin conductance across the day may provide some insight into changes in emotional arousal before and after LOC without explicitly drawing a person's attention to their emotional state. Emerging research also suggests that it may be possible to detect some changes in mood through passive monitoring of other aspects of a person's behavior, such as the speed or pitch of their speech [which can be recorded throughout the day using cellphones (118)].

A second limitation relates to who has been included in experience sampling research on LOC. The vast majority of studies, particularly studies with clinical populations, have only included women. Among studies that have included men, none have examined whether associations between NA and LOC differ across gender. Relatedly, most studies have used participant samples that are predominantly (often over 95%) white. This is despite the fact that LOC is equally or more common among people of color than among white individuals (119). Additional research is therefore urgently needed to examine whether associations between NA and LOC are similar across gender and race/ethnicity. Differences across ethnicity are particularly important to investigate further because some factors that moderate NA-LOC eating associations, such as eating expectancies, may be based in sociocultural norms that could vary across diverse groups.

A third limitation is that experience sampling research on NA-LOC relationships has thus far paid little attention to biological and environmental contextual factors that might impact NA, LOC, or their association at a given moment in time. Emotions, and adaptive responses to emotions, are intimately tied to the context in which they occur (120). Both environmental and biological factors can influence the intensity and persistence of negative emotions, and may also impact the extent to which they lead to LOC. To date, the vast majority of experience sampling research on LOC has paid little attention to the impact of a person's momentary environment (e.g., whether someone is alone or with others, the presence of palatable food or messages stigmatizing larger bodies) on nuances of emotions or behavior. Relatedly, almost no research has examined the biological context in which negative emotions and LOC occur. Biological factors that may be outside of a person's immediate awareness (e.g., menstrual cycle phase, blood sugar levels, fatigue) may increase NA, or reactivity to negative events, in a way that increases risk for LOC. A more detailed understanding of the moment-to-moment environmental and biological factors that may increase NA or strengthen NA-LOC associations is needed to refine and clarify affect regulation models. For example, understanding the impact of biological and environmental moderators on NA-LOC associations could inform our understanding of why LOC might persist even if it is not immediately reinforcing on average (i.e., whether it is intermittently reinforcing based on the context). Experience sampling research is uniquely poised to collect information about momentary biological and environmental factors before, during, and after LOC in ecologically valid settings. When combined with other minimally invasive measures (e.g., saliva collection for hormone levels, heart rate monitoring), it has the potential to help us identify with precision when and why NA may trigger maladaptive eating behavior. In the future, additional research in other cultural contexts could also help illuminate the influence of broader cultural factors and beliefs on NA-LOC relationships, and ensure that research to date generalizes to diverse populations.

A proposed, revised affect-focused model of LOC is depicted in Figure 2. Figure 2 reflects existing research showing that LOC is proceeded by an increase in NA, and followed by an immediate increase and gradual decline in NA. The dashed line during the period of LOC represents uncertainty regarding what happens to emotions while LOC is in progress, and the need for additional creative research to examine this phase (e.g., using passive monitoring of heart rate or vocal tone). While overall NA is shown to decrease in the hours following LOC, a separate line is depicted for guilt/shame, which evidence suggests remains elevated even after other negative emotions begin to subside. This nomothetic pattern of NA surrounding LOC is qualified by several layers of biological and environmental factors that may either directly increase NA, or alter the strength of NA-LOC associations. These factors include a person's momentary biological state, trait-level individual differences (e.g., negative urgency, beliefs about the effects of eating on emotions), momentary proximal environmental factors, and the broader cultural context. Which of these factors most strongly influence the pattern of effects shown in the central graph in Figure 2, or how they might interact with each other, represents an exciting area for future research.

Of note, the proposed affect-focused model in Figure 2 shares some similarities with conceptualizations of other forms of psychopathology. Most notably, it is similar to models of non-suicidal self-injury (NSSI), in which negative emotions may drive a person to injure themselves with (in many cases) the ultimate goal of feeling better. As with LOC, people frequently report engaging in NSSI to relieve NA (121). However, evidence from experience sampling research is mixed with regards to whether NA actually decreases following NSSI in daily life (122), with some research finding that NA peaks shortly after NSSI and slowly decreases over subsequent hours (123). As with LOC, motivations for engaging in NSSI may be complex, with elements of wanting to feel better (e.g., release difficult emotions) and wanting to feel worse (e.g., punish self for not meeting one's standards) (121). Engaging in NSSI may also increase PA, similar to consumption of palatable food in LOC (124). These similarities suggest that similar underlying mechanisms may be at play with respect to NSSI and LOC, whereby difficulty tolerating and managing distress leads to maladaptive behavior in a desire to feel better (even if this behavior is not always effective, and there are negative longer term consequences).

Conceptual similarities between the proposed affect-focused model of LOC and NSSI suggest that similar treatment approaches may be appropriate. In particular, DBT has shown promise for both NSSI and disorders characterized by LOC by targeting underlying difficulties with distress tolerance (i.e., the ability to experience NA without engaging in maladaptive behavior), problem solving, and emotion regulation (125). However, while LOC may function similarly to NSSI with respect to emotion regulation, it may also differ in some important respects. In particular, LOC may be propelled not only by NA, but also by biological drives to consume food after food restriction. LOC may therefore require additional treatment elements beyond simply improving distress tolerance and emotion regulation, such as implementing regular, balanced meals, as occurs in CBT for EDs (4). Because consumption of palatable foods during LOC may be pleasurable, it may also be necessary to incorporate therapeutic elements to increase experience of PA, rather than just decreasing or improving tolerance for NA.

While additional work remains to be done, research to date nevertheless offers some practical clinical implications. Evidence that elevated NA increases risk for subsequent LOC supports key aspects of existing CBT-based treatments, such as working with clients to identify the situations that are most likely to trigger NA and develop strategies to help regulate their emotions at these times (4). Adding DBT-based techniques from protocols adapted for BED and BN (20) that teach distress tolerance (e.g., distraction, self-soothing) and problem solving (e.g., weighing pros and cons) skills may have added benefit in helping individuals develop the tools necessary to abstain from LOC when feeling negative emotions. Clients may also benefit from behavioral experiments that challenge the belief that LOC decreases NA. For example, clients could be asked to record how they feel immediately before and after LOC, and these data could be used to challenge ideas about the emotion regulatory properties of LOC. Finally, mindfulness-based exercises and emotional exposures [i.e., exercises to build acceptance and tolerance of emotions, as described in Barlow's Unified Protocol (126)] could be used to help decrease reactivity to NA and help clients learn that negative emotions decrease on their own, even without LOC or other maladaptive behaviors intended to numb these feelings. Exposure and response prevention (ERP) for obsessive-compulsive disorder (127) offers some complementary techniques. In traditional ERP for OCD, clients are coached to refrain from engaging in compulsions when experiencing obsessive thoughts so that they may learn that compulsions are not necessary for reducing anxiety and avoiding catastrophic outcomes. Analogously, clients with LOC could be intentionally exposed to situations (real or imaginal) that provoke NA, and coached to refrain from engaging in maladaptive eating behavior to learn that this behavior is not necessary for feeling better. Similar techniques have already been effectively incorporated to some extent in existing cue exposure and response prevention treatments for LOC (128), though the focus has more often been on exposure to environmental factors associated with LOC (e.g., seeing a particular food, being in a particular room in the house) than affective factors that may trigger this behavior.

Though the standard affect regulation model may require some revision, experience sampling research has affirmed the importance of NA as a driver of LOC in daily life. Future research that incorporates individual differences and momentary biological and environmental factors will help us continue to expand our understanding of when, why, and for whom negative feelings may trigger dysregulated eating.

The author confirms being the sole contributor of this work and has approved it for publication.

This work was supported by a grant from the National Institute of Mental Health (R01 MH111715-03S1). The content is solely the responsibility of the author and does not necessarily represent the official views of the National Institute of Mental Health.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.