The Potential Application of Event-Related Potentials (ERPs) to Enhance Research on Reward Processes in Eating Disorders

Abstract

Reward-related processes have been posited as key mechanisms underlying the onset and persistence of eating disorders, prompting a growing body of research in this area. Existing studies have primarily utilized self-report, behavioral, and fMRI measures to interrogate reward among individuals with eating disorders. However, limitations inherent in each of these methods (e.g., poor temporal resolution) may obscure distinct neurocognitive reward processes, potentially contributing to underdeveloped models of reward dysfunction within the eating disorders. The temporal precision of event-related potentials (ERPs), derived from electroencephalography (EEG), may thus offer a powerful complementary tool for elucidating the neurocognitive underpinnings of reward. Indeed, a considerable amount of research in other domains of psychopathology (e.g., depression, substance use disorders), as well as studies investigating food reward among non-clinical samples, highlight the utility of ERPs for probing reward processes. However, no study to date has utilized ERPs to directly examine reward functioning in the eating disorders. In this paper, we review evidence underscoring the clinical utility of ERP measures of reward, as well as a variety of reward-related tasks that can be used to elicit specific ERP components with demonstrated relevance to reward processing. We then consider the ways in which these tasks/components may be used to help answer a variety of open questions within the eating disorders literature on reward. Given the promise of ERP measures of reward to the field of eating disorders, we ultimately hope to spur and guide research in this currently neglected area.

Approximately 28.8 million people living in the United States will experience an eating disorder (i.e., anorexia nervosa [AN], bulimia nervosa [BN], and binge-eating disorder [BED]) within their lifetime.¹ These serious psychiatric disorders are associated with significant psychological and medical comorbidity, psychosocial impairment, chronicity, and mortality.² Unfortunately, existing interventions for eating disorders are often unsuccessful in achieving lasting remission,^3–5 highlighting the need for a more nuanced understanding of associated risk and maintenance factors to help guide the development of improved interventions.

Eating disorders are characterized, in part, by the over- or under-consumption of food, which is a primary reinforcer, or a substance that is naturally rewarding and fulfills a biological need. Because of this, dysregulation in reward-related processes have been posited to underlie eating disorder onset and persistence. For example, some researchers suggest that disorders characterized by loss of control and overeating (e.g., AN-binge/purge type, BN, BED) may be partially attributable to an exaggerated reward response when consuming food, contributing to an increased motivational salience of food (i.e., an increased anticipatory drive for food due to the learned association between food and reward).^6,7 Alternatively, individuals with these binge-type eating disorders may experience a diminished reward response during food consumption, resulting in increased food intake to compensate for the experienced reward deficits.⁸ Among individuals with restrictive eating disorders (e.g., AN-restricting type), reduced reward responding to food has been theorized to occur alongside elevated responsivity to rewards associated with dietary restriction and physical activity, resulting in a pathological pattern of reduced food intake and compulsive exercise.^9–12

Given the hypothesized role of reward-related processes in the eating disorders, a substantial body of research using self-report, behavioral, and neuroimaging techniques to investigate potential reward abnormalities among individuals with eating disorders has emerged. Although findings from self-report and behavioral data are mixed, a growing number of studies support an association between eating disorders and altered reward-responding.^13–18 Research using functional magnetic resonance imaging (fMRI) has recently gained momentum, but these studies too have frequently produced inconsistent or contradictory results.^17,19 For example, research in AN, BN, and BED has sometimes indicated enhanced activation of reward circuitry in response to palatable food cues,^20–23 and at other times, reduced activation.^8,24–26

Mixed findings in the fMRI literature may in part reflect substantial methodological differences between studies.^17,19 However, the relatively limited temporal resolution of fMRI may also contribute to mixed results. Although the excellent spatial resolution of fMRI allows researchers to pinpoint the neural systems involved during reward processing, the changes in blood oxygenation levels (i.e., the BOLD signal) that are typically measured with fMRI resolve over a period of a few seconds to tens of seconds.²⁷ As many reward-related cognitive processes occur on the timescale of milliseconds,²⁸ fMRI, as well as self-report and behavioral measures, may be unable to distinguish between important neurocognitive processes involved in reward responding. By obscuring distinct neurocognitive processes, a sole reliance on these forms of data may contribute to underdeveloped models of reward dysfunction within the eating disorders, and could ultimately hamper targeted interventions.

Electroencephalography (EEG) may therefore provide a powerful complement to self-report, behavioral, and fMRI data in the field’s effort to clarify the role of reward-related processes within eating disorders. EEG is a relatively low-cost, non-invasive procedure that involves recording electrical activity produced by neurons in the brain through electrodes placed on the scalp.²⁹ Critically, EEG possesses high temporal resolution, capturing this neural activity nearly instantaneously. When the EEG record is time-locked to an event of interest (e.g., the receipt of a reward), a series of systematic voltage changes (i.e., positive and negative deflections) create a patterned waveform of voltage over time, and this waveform is called the event-related potential (ERP). ERPs capture multiple signals, called ERP components, that can be used to distinguish between distinct neurocognitive processes occurring within milliseconds of one another.³⁰

Within the eating disorders literature, ERPs have primarily been used to examine altered attention or inhibitory control related to food and body image cues.^31–34 Although this research base is small, results suggest that EEG may offer unique insights into the neurocognitive processes underlying eating disorders. For example, accumulating ERP evidence suggests that binge-type eating disorders are associated with attention biases to food at multiple early processing stages.³² These ERP findings further indicate that the biases may be enhanced by environmental cues³⁵ and persist even when the appetitive value of food is experimentally reduced.³⁶

While no study to date has utilized ERPs to study reward-specific processing within the eating disorders, initial research among healthy adults indicates that ERPs may allow for fine-grained investigations of the relationship between food reward processes and eating behavior.³⁷ Moreover, research in other clinical domains suggests that ERP markers of reward offer powerful tools for understanding disease risk and progression. This literature is particularly well-developed in the field of depression, where reward-related ERPs have been shown to prospectively predict depression onset and treatment response, even when controlling for other known risk/maintenance factors.^38–41

Given the theoretical relevance of reward to eating pathology, unique advantages of EEG over other forms of data-collection, and demonstrated clinical utility of reward-related ERP components in other psychiatric disorders, we propose that the use of ERPs to study reward functioning among individuals with eating disorders is an important and underexplored area. To better facilitate research in this vein, we next review several relevant paradigms and associated tasks that may be used to elicit neural reward responding, as well as specific ERP components related to reward processing.

ERP Reward Paradigms and Associated Tasks

A variety of tasks are regularly implemented to study reward processes using ERPs. These tasks attempt to isolate unique stages of behavior and cognition, allowing researchers to manipulate key variables of interest and identify distinct reward processes that may be impacted. For example, many tasks allow researchers to distinguish between anticipatory reward processes (which occur before reward receipt) and outcome reward processes (which occur after reward receipt). We have broadly categorized these tasks into three basic paradigms, or general task structures: (a) passive reward tasks, (b) instrumental reward tasks, and (c) reward learning tasks.⁴² However, it should be noted that these categories share considerable overlap. Further, although we aim to highlight some of the most commonly used reward tasks, this list cannot be considered exhaustive. Finally, these tasks are often adapted from (or similar to) common fMRI tasks, allowing for ready integration of results with fMRI literature. However, it is important to note that differences between EEG and fMRI often require differences in task features (e.g., timing, stimulus properties, trial numbers). Therefore, care must be taken when adapting tasks for ERPs.³⁰

Passive Reward Paradigms

The simplest paradigm involves the passive presentation of rewards, in the absence of any active effort by the participant to obtain the rewards.^43,44 For example, in the slot machine task, three stimuli (e.g., kiwi, pear, cherries) are presented consecutively, and participants win a reward (e.g., money) when all three presented stimuli are identical (e.g., three cherries). In the most basic version of the task, there is no probabilistic relationship between the stimuli, meaning that participants are unable to predict when rewards might occur. However, researchers have also developed passive reward tasks with probabilistic relationships between consecutive stimuli,^20,45 which allows for the examination of passive reward processing elicited by stimuli that predict likely upcoming rewards, as well as the comparison of responses to expected and unexpected reward outcomes. This is noteworthy given the important role of expectancies in models of reward processing.^46,47

Instrumental Reward Paradigms

In contrast with passive paradigms, instrumental reward paradigms require participants to complete an action correctly in order to obtain the reward. An increasingly popular instrumental reward task is the monetary incentive delay task (MID).⁴⁸ ERP versions of the MID,⁴⁹ and similar tasks,⁵⁰ can be used to distinguish at least five discrete stages of reward processing. In the first (i.e., cue) stage, an incentive cue is presented that indicates which of two types of trials the upcoming trial will be: in reward trials, rewards will be won or lost based on subsequent behavioral performance, while in non-reward trials nothing will be won or lost regardless of behavioral performance. In the second (i.e., preparation) stage, a brief delay period provides the participant with an opportunity to prepare for an upcoming test stimulus and response. In the third (i.e., cognitive/motor) stage, the test stimulus is presented and the participant must accurately execute a response (e.g., a rapid button press). In the fourth (i.e., outcome anticipation) stage, participants are left to anticipate the outcome of their action during a second brief delay. Finally, in the fifth (i.e., feedback) stage, the outcome is revealed (e.g., rewards were won, lost, or unaffected). ERP activity can be time locked to the incentive cue, test stimulus, behavioral response, and feedback to examine different anticipatory- and outcome-related reward processes.

One of the most frequently used instrumental reward tasks in clinical research is a guessing or gambling task called the doors task.^51–53 Although there are many variations of this task, its core structure is simple: Participants repeatedly choose between two doors presented on a screen, attempting to pick the door that will result in receipt of rewards (typically money) and avoid the door that will result in loss of rewards. After each choice, participants receive outcome feedback. ERP activity is time-locked to the feedback, separately for wins and losses, in order to compare the neural responses elicited by rewarding and punishing outcomes. In the doors task, there is typically no fixed or probabilistic relationship between the door selected and the outcome received – win and loss outcomes are random and equally likely. This has the advantage of controlling for the likelihood of gains and losses; however, it also means that real learning about which door is more advantageous to choose cannot occur.

Reward Learning Paradigms

Variations of instrumental reward tasks can also be used to study reward learning. In their simplest form, reward learning tasks are similar to the doors task, except that the different choice options (e.g., left door vs. right door) are rewarded based on different reinforcement schedules (e.g., the left door leads to a gain outcome 70% of the time, and the right door 30% of the time). ERP activity is typically time-locked to the feedback stimuli indicating gains and losses, and can be examined in relation to behavioral aspects of learning on the task (e.g., the degree to which participants develop a bias towards heavily rewarded response options). A variety of reward learning tasks have been utilized in ERP research in which underlying, usually probabilistic, relationships between behaviors and reward outcomes can be learned,^49,54–58 allowing researchers to isolate the neural activity associated with behavioral performance, as well as with computationally-modeled learning parameters.

ERP Reward Components

Each event in the tasks described above (e.g., incentive cue, behavioral response, outcome feedback) elicits a series of ERP components that index distinct neurocognitive reward processes. These components are typically identified by their scalp location (e.g., frontal or posterior), timing (e.g., 200 – 300 ms after the event), polarity (positive or negative voltage), and eliciting conditions (e.g., an unexpected outcome). In this section, we will review several of the most commonly utilized ERP components for studying reward processing in the context of clinical research. We begin with two components related to reward outcome processing, as these components have received the most attention in the literature to date. Then, we review three components relevant to anticipatory reward, which is an area of increasing interest within the eating disorders field. Please see Figure 1 for a representation of each component and its relation to task events. Note that this selection of reward ERP components is not exhaustive. Further, the ERP technique represents only one way to utilize EEG data. Time-frequency analysis, for example, can be used to help disentangle overlapping reward-related EEG activity that occurs at different frequencies (e.g., reward and loss related activity in the delta and theta frequency bands, respectively), representing distinct functional processes generated by different brain regions.⁵⁹ For a broader review of EEG methods to assess reward, see Glazer et al. (2018)²⁸.

Figure 1.

Example Monetary Incentive Delay Task (MID) with Associated Reward-Related Event-Related Potential (ERP) Components

Note. Multiple ERP components, sensitive to distinct neurocognitive reward processes, can be elicited within a single task. In this example (modeled on the MID task), the occurrence of an incentive cue signifies that behavioral performance in the upcoming trial will be rewarded/punished with monetary gain/loss, while a neutral cue signifies that no rewards/punishments are possible. After the behavior is completed, participants await and then receive feedback indicating the monetary outcome of their behavior. Three ERP panels, highlighting five ERP components, are represented at the top of the figure. Each ERP is time locked to the onset of a given task event (in this example, the cue stimulus or the feedback stimulus), which is represented as 0 milliseconds (ms) on the x-axis. By convention, positive voltage is plotted downward on the y-axes and negative voltage is plotted upward on the y-axes. Note that in this example the SPN component is time locked to the feedback, which it occurs prior to, thus the x-axis for the second ERP panel shows time leading up to the feedback event occurring at 0 ms. At the bottom of the figure, a brief, simplified description of each component is provided, and its place within the continuum of anticipation to outcome processing is indicated. For empirical ERPs and relevant task details, see, for example, Novak et al., 2016.¹¹⁶

Reward Outcome: The Reward Positivity (RewP)

The reward positivity (RewP, also known as the feedback-related negativity or FRN) is a positive ERP observed at frontocentral electrodes that reflects an early neural response to rewarding outcomes. The RewP occurs rapidly, approximately 200 – 350 ms after outcome feedback, and appears to reflect an initial outcome evaluation process that differentiates favorable and unfavorable outcomes. The component is typically calculated by subtracting the ERP activity elicited by non-rewards (i.e., losses or missed opportunities for gain; red waveform in Figure 1) from the ERP activity elicited by rewards (green waveform in Figure 1).⁵³

The RewP is thought to result from phasic increases in midbrain dopamine following rewarding outcomes, and is associated with activity in reward-related brain regions including the striatum, medial prefrontal cortex, and anterior cingulate cortex.^60–64 Functionally, the reward positivity may support reward-learning processes by signaling that the outcome of a given behavior was better than expected (i.e., a positive reward prediction error), and thus that the behavior should be more highly valued going forward.^60,65 In support of this, the RewP is typically sensitive to both the magnitude and expected likelihood of reward, such that larger and more unexpected rewards elicit a greater RewP.⁶⁶ Given the theorized relevance of the RewP to reward learning processes, it may be especially valuable within reward learning paradigms, though it can be elicited in any paradigm involving the opportunity to win rewards, including, albeit to a lesser extent, in passive reward tasks.⁶⁶

While the exact nature of the RewP is still a matter of some debate, there is increasing evidence from clinical and non-clinical populations that the component provides a reliable and objective measure of an individual’s sensitivity to reward.^53,67,68 For example, RewP amplitude is related to individual differences in reward sensitivity as measured by both self-report and behavioral indices, even when the self-report and behavioral indices are uncorrelated, suggesting that the RewP may tap a fundamental aspect of reward processing.⁶⁹ Further, a growing body of evidence indicates that individuals with major depressive disorder (a condition characterized by low positive affect and/or anhedonia)⁷⁰ exhibit a reduced RewP following reward receipt, when compared with healthy controls.^71,72 Similarly, accruing evidence suggests RewP responses are also altered among other populations known to have aberrant reward sensitivity, such as those with substance use disorders^73,74 and Parkinson’s disease.⁷⁵ Importantly, a blunted RewP appears to represent a unique risk factor for the onset of depression in children and adolescents^39–41 and predicts enhanced treatment outcomes among depressed individuals.³⁸ As emerging research suggests that the RewP may be a malleable target for clinical interventions involving psychotherapeutic⁷⁶ and neuromodulation techniques,^77,78 interrogation of the RewP within the eating disorders may be a fruitful avenue with meaningful clinical implications. However, to our knowledge, no study has yet systematically examined the reward positivity among individuals with eating pathology (though a small number of studies have examined related ERP activity^{79 80}).

Reward Outcome: The Feedback-P300 (FB-P300)

The P300 is a positive ERP that is elicited by motivationally significant stimuli (i.e., stimuli that activate approach, avoidance, and learning systems in the brain).^81,82 For example, an intrinsically arousing image (e.g., a snake or an erotic photograph) will elicit a P300, but so will a neutral stimulus (e.g., a 500 Hz tone) if one’s current goal is to press a button every time that stimulus occurs. In each case, a motivationally significant event has occurred, and a cascade of neural processes are initiated to guide immediate and future behavior. As part of this cascade, the P300 is thought to most closely reflect attentional and memory-related processes that guide future goal-directed behavior, such as the process of updating a mental model of the environment in order to accommodate the significant event and shape future expectations about the environment.⁸² The P300 is typically maximal at centroparietal electrodes approximately 300 – 600 ms after stimulus onset, and results from widespread cortical activity, which may in part be coordinated by the locus coeruleus noradrenergic system.^83–85

Not surprisingly, a P300 is elicited by feedback signifying a rewarding outcome. In this case, it is referred to as the feedback-P300 (FB-P300), and it follows immediately after the RewP. Like the RewP, the FB-P300 tends to vary with the magnitude and probability of reward outcomes.⁸⁶ Unlike the RewP, however, it does not index the valence of an outcome – it occurs in response to both positive and negative outcome feedback (e.g., wins and losses), as both outcomes are motivationally significant. The RewP and FB-P300 may thus be useful in concert as distinct measures of reward processing, reflecting the rapid evaluation of valence (How good is the outcome?) and significance (How important is the outcome?), respectively.^87,88

A large body of work illustrates the utility of the P300 within clinical research^89,90, and recent work indicates that the FB-P300 elicited in reward paradigms may have similar value. For example, a blunted FB-P300 to reward outcomes is related to increased risk for current and future depression⁹¹ and attention-deficit/hyperactivity disorder (ADHD)⁹² in children and adolescents. Consistent with the functional distinctions outlined above, the FB-P300 and RewP appear to tap independent aspects of reward-related dysfunction in clinical samples, and demonstrate distinct patterns of association with unique manifestations of psychopathology, potentially signaling differential intervention targets.^91–95

While the P300 has been used fruitfully in research on eating disorders,^31,32,34 no studies on eating disorders have, to our knowledge, utilized reward-specific paradigms that elicit the FB-P300 to reward outcomes. Of particular relevance, however, is a recent study, which examined the FB-P300 elicited by food rewards in a sample of individuals without a history of psychiatric disorders.³⁷ Participants in this study completed a doors task in which images of low, medium, and highly preferred snack foods were revealed after selecting one of three possible doors. Participants were told that they would be able to consume the snack food that was most frequently revealed at the end of the task. The authors found that the FB-P300 scaled according to preference when participants were hungry. However, when participants were in a satiated state with regard to their preferred food (i.e., after eating their highly preferred food until they no longer wanted any more), the FB-P300 no longer differentiated high and medium preference foods. These results confirm that the FB-P300 is a meaningful index of the subjective motivational significance of specific foods, which is sensitive to devaluation based on motivational state.³⁷ In contrast, while the RewP also scaled with participant preference in this study, it was less sensitive to the devaluation manipulation, further highlighting the utility of ERPs for disentangling distinct reward processes occurring on a rapid timescale.

Reward Anticipation: The Cue-P300, CNV, and SPN

ERP reward paradigms have traditionally focused on reward-outcome processing,⁹⁶ but recent work has broadened this focus to examine anticipatory aspects of reward. For example, in tasks that utilize a cue to indicate whether there is a potential to win or lose rewards based on subsequent behavioral performance (e.g., MID task), this cue elicits a P300, which varies with the motivational salience or value of the cue (i.e., whether the cue indicates potential for a small, large, or no reward).^97–100 The P300 to the cue, referred to as the cue-P300, reflects an early motivational response to the prospect of reward, rather than to its receipt.

Following the cue-P300, a slow, ramp-like negative ERP called the contingent negative variation (CNV) often develops over central electrodes, leading up to an expected stimulus that will require a speeded response in order to gain (or avoid losing) a reward.^101,102 The CNV is thought to reflect motor and attention-related preparation in order to facilitate behavioral performance, and increases in magnitude with increasing potential reward value.^91,92 This effect is associated with activity in the supplementary motor area, thalamus, prefrontal cortex, and ventral striatum.^103,104 Therefore, while the cue-P300 reflects the initial reaction to reward prospect, the CNV reflects subsequent, and more sustained, preparedness for behavioral action in pursuit of reward.

Finally, after a behavioral response is made, and a participant is awaiting feedback about the outcome of the behavior, a negative ERP called the stimulus preceding negativity (SPN) typically occurs over frontocentral electrodes in the 200 ms prior to feedback.¹⁰⁵ The SPN is thought to reflect attentional resource mobilization in anticipation of feedback, which is enhanced when the upcoming feedback will signify the gain or loss of reward (relative to feedback unrelated to reward).¹⁰⁶ This effect is linked to the dopaminergic reward system and is closely associated with activity in the anterior insula.^28,107 Thus, while the CNV reflects motivated preparation to act in the pursuit of reward, the SPN likely reflects motivated anticipation of the outcome of this action.

Variations in the reward cue-P300, CNV, and SPN have all been associated with reward-related mental health concerns, including blunted anticipatory activity related to depression,^{91,94,108,109} history of suicide attempts,¹¹⁰ and ADHD,¹¹¹ but amplified anticipatory ctivity related to substance use disorder¹¹² (though see Zhao et al., 2017¹¹³). These anticipatory components have also been shown to predict therapy compliance, further supporting their clinical utility.¹¹⁴ Importantly, the cue-P300, CNV, and SPN appear to differentiate distinct aspects of dysfunctional reward anticipation.^115,116 For example, individuals with methamphetamine use disorder who completed a version of the MID were more sensitive to cues (cue-P300) signaling potential reward as well as cues signaling potential punishment, compared to healthy controls.¹¹⁷ However, when anticipating the outcome of their behaviors (SPN), these individuals showed bias towards greater anticipation of potential rewards only. Thus, these distinct anticipatory ERP components suggest that individuals with substance use disorder may not simply have a hyper-sensitive anticipatory reward response. Instead, they may be initially highly sensitive to both the potential rewards and potential punishments of a given behavior, and are only later biased towards potential rewards when awaiting the consequences of a behavior. To our knowledge, no studies have examined the relationship between eating disorders and the cue-P300, CNV, or SPN in the context of reward paradigms. However, in a particularly relevant study, Versace et al. (2017)¹¹⁸ found that healthy individuals with a stronger cue-P300 like response⁸¹ to food images that predicted the receipt of candy, relative to erotic images, ate twice as much of the candy compared to those who had the opposite ERP pattern, suggesting that anticipatory reward ERPs elicited by food cues are meaningfully associated with eating behavior.

Discussion

Numerous authors have highlighted the theoretical relevance of reward to the onset and persistence of eating disorders,^6,8–12 and initial data using self-report, behavioral, and fMRI measures provide some support for these assertions. However, limitations in these data-collection methods may obscure discrete reward-related neurocognitive processes, contributing to mixed findings and underdeveloped models of eating pathology. We propose that ERPs offer a meaningful complement to this work by allowing researchers to parse apart discrete aspects of reward that may otherwise be overlooked, and to differentiate these reward-related processes from other relevant perceptual, attentional, and evaluative processes.¹¹⁹ We expand upon this proposal below, and provide additional suggestions for reward ERP approaches to unanswered eating-disorders questions in Table 1.

Table 1.

Reward in eating disorders: Research questions and suggested ERP approaches

Questions	ERP Approaches
Are eating disorders associated with altered reward processing? Which reward processes are altered in eating disorders? Are different eating disorder diagnoses associated with different patterns of reward responsivity? Do these patterns depend on the types of rewards? Do alterations in reward neural activity relate to task behavior?	Identify which ERP components distinguish different eating disorders from healthy controls and from each other. Clarify whether abnormal reward responding is specific to certain stimuli by varying the types of available rewards (e.g., high- and low-calorie foods, food restriction opportunities). Test how reward ERPs are associated with task performance (e.g., learning, memory, reaction time).
How do reward processes relate to real-world disordered eating behavior? Do individual differences in reward processing relate to eating behavior in a laboratory setting? Do individual differences in reward processing relate to differences in eating disorder behavior outside of the laboratory? Do individual differences in reward processing interact with other factors to predict eating behavior?	Examine the association between reward ERPs and eating behavior in a lab-based test meal. Examine association between reward ERPs and eating disorder behaviors in daily life (e.g., using ecological momentary assessment). Examine how reward ERPs moderate the relationships between momentary state/environment (e.g., exposure to food advertisements, stressors, or emotional events) and eating disorder behavior.
Are there reward-related biomarkers for eating disorder risk or treatment response? Can reward ERPs predict future eating disorder onset or severity? Can reward ERPs predict treatment response or compliance? Can reward ERPs provide targets for intervention?	Use reward ERPs as baseline predictors of longitudinally assessed eating disorder development in children and adolescents. Use reward ERPs as baseline predictors of change in eating disorder symptoms following treatment. Use changes in reward ERPs (e.g., pre-, mid-, post-treatment) to measure the effectiveness of treatments aimed at changing basic reward processing.

At the simplest level, ERPs can be used to help provide an answer to the question of when, within the streams of neurocognitive activity surrounding reward-related events, processing is altered among individuals with eating disorders. For example, ERPs can support ongoing efforts using self-report, behavioral, and fMRI approaches to identify whether individuals with eating disorders are more (or less) sensitive to the prospect of reward, to the receipt of reward, or to both.⁸ In addition, ERPs can facilitate a more fine-grained analysis of these temporal dynamics. For example, if individuals with eating disorders do show heightened sensitivity during the reward anticipation stage, ERPs could be used to determine the extent to which that sensitivity reflects an initial reaction to environmental cues (cue-P300) or the subsequent mobilization of cognitive and motor resources in support of action to obtain the reward (CNV).

Importantly, ERPs can also be used to help identify what precise reward processes are affected among individuals with eating disorders. For example, ERPs could be used to clarify the extent to which reward abnormalities are specific to disorder-relevant stimuli (e.g., food, restriction, exercise), or may represent more global dysfunction in reward processing. Current evidence suggests that the reward value of food (generally measured as the amount of time, effort, or money an individual will expend to obtain food) may be altered among individuals with eating disorders,¹⁵ and more recent behavioral and fMRI work has sought to identify what processes relevant to the computation, storage, or utilization of food reward value may contribute to these abnormalities.^120,121 As distinct ERP components may provide dissociable indices of reward value, (with the RewP potentially providing a more basic measure of food reward value and the FB-P300 providing a more context or goal dependent measure of value³⁷) ERPs may provide a novel approach to examining the relative subjective value of multiple disorder-relevant rewards, while also identifying what aspect of reward valuation processing is more strongly affected by the disorder.

ERPs can also provide mechanistic insight into how reward processes, such as reward learning, may lead to or maintain disordered eating behavior.¹⁶ For example, researchers have recently sought to understand whether maladaptive eating behaviors result from impaired goal-directed learning mechanisms, or from deficits in more automatic or habitual forms of learning.^122–124 ERPs can be used to monitor learning and memory processes as they are happening, and recent research suggests that ERPs can specifically differentiate the learning mechanisms that underlie goal-directed and habitual behavior (i.e., model-based and model-free learning, respectively).^125,126

Finally, reward ERPs could potentially provide biomarkers of eating disorder risk and treatment outcomes, as well as direct targets for intervention.¹¹⁹ For example, the RewP has been shown to discriminate between other psychological disorders and to uniquely predict disorder onset, remission, and response to treatment.⁵³ Further, the RewP may be malleable through psychotherapeutic and neuromodulation techniques,^76–78 opening a potential pathway for influencing the biological mechanisms underlying eating pathology.

In sum, ERPs have the potential to provide an important complement to ongoing research on reward in the eating disorders, but have been almost entirely neglected in this capacity to date. In this paper, we have attempted to outline a variety of tasks and ERP components that may be used to support research in this area. Notably, given the intrinsic limitations of individual forms of data-collection (e.g., low spatial resolution in EEG, low temporal resolution in fMRI, retrospective recall bias in self-report measures), we strongly encourage eating disorder researchers to consider multi-method approaches to studying reward in this population. For example, ERPs can be combined with fMRI in order to characterize the temporal dynamics of reward processing within well-defined neural networks, potentially allowing for a much more comprehensive understanding of how these reward processes are implemented by the brain, as well as how they may be dysregulated.¹²⁷ In addition, ERP measures of reward can be used in concert with naturalistically-assessed reward measures (e.g., captured via ecological momentary assessment) to better understand how neurocognitive markers of reward may relate to real-world behaviors. Ultimately, we contend that leveraging the numerous strengths of EEG/ERPs (ideally in combination with other measures of reward responding) will provide a powerful tool for testing theoretical models of reward in eating disorders, and isolating specific targets for intervention.

Public Significance Statement.

Abnormalities in reward functioning appear to contribute to eating disorders. Event-related potentials (ERPs) offer temporally precise measures of neurocognitive reward processing, and thus may be important tools for understanding the relationship between reward and disordered eating. However, research in this area is currently lacking. This paper attempts to facilitate the use of ERPs to study reward among individuals with eating disorders by reviewing the relevant theories and methods.