Contra‐Hedonic Functions of Self‐Destructive Behaviors in Adolescents

ABSTRACT

Objective

Many self‐destructive behaviors, including eating disorder (ED) behaviors (restrictive eating, binge eating, and purging), and nonsuicidal self‐injury (NSSI), serve an emotion regulation function. To date, it has been assumed that this is pro‐hedonic in nature (e.g., to decrease negative affect). However, contra‐hedonic motivations (to increase/maintain negative affect, decrease positive affect) are prevalent during adolescence and may also reinforce self‐destructive behaviors. In this study, we explored the presence, frequency, and clinical characteristics of contra‐hedonic functions of ED behaviors and NSSI among adolescents.

Method

Participants were drawn from a larger longitudinal study and recruited online via social media. The present study included a transdiagnostic sample of 137 adolescents (75.91% White, 9.0% Hispanic, 75.18% female, mean age = 15.10 years) who completed surveys assessing the frequency, severity, contra‐hedonic, and pro‐hedonic functions of ED behaviors and NSSI.

Results

Adolescents reported engaging in restrictive eating, binge eating, purging, and NSSI for each contra‐hedonic function assessed, with 81% of participants endorsing at least one contra‐hedonic function of ED behaviors and/or NSSI. Contra‐hedonic motivations were highest for NSSI compared to ED behaviors and were consistently associated with a desire to self‐punish.

Conclusions

This study sheds light on a largely unexplored dimension of reinforcement functions of ED behaviors and NSSI: adolescents may not only engage in these behaviors to feel better, but also to make themselves feel worse. Further characterization of both pro‐ and contra‐hedonic functions of maladaptive behaviors in adolescents represents a promising avenue through which behavioral interventions can be more precisely designed and targeted.

Summary

• To date, it has been assumed that self‐destructive behaviors, such as eating disorder behaviors and nonsuicidal self‐injury, serve a pro‐hedonic emotion regulation function—that is, to increase positive affect or decrease negative affect.

• This study explores whether adolescents engage in eating disorder behaviors and non‐suicidal self‐injury for contra‐hedonic motivations—that is, to increase/maintain negative affect or decrease positive affect.

• We found that contra‐hedonic motivations for these behaviors were not only present but prevalent.

1. Introduction

Eating disorder (ED) behaviors (restrictive eating, binge eating, and purging) and nonsuicidal self‐injury (NSSI) are highly comorbid and follow a similar developmental trajectory, prompting researchers to investigate shared underlying factors (Claes and Muehlenkamp 2014). One unifying explanation is these behaviors serve an emotion regulation function (e.g., to reduce negative emotions/increase positive emotions) (Nock and Prinstein 2004; Wang et al. 2021; Wedig and Nock 2010).

It has historically been assumed these emotion regulation functions are pro‐hedonic: that people engage in ED behaviors and NSSI to feel better. However, even healthy people pursue activities inducing negative emotions (e.g., listening to sad music, watching scary movies). Therefore, emotion regulation models of self‐destructive behaviors may need to incorporate the fact that sometimes people want to feel worse, and some behaviors are contra‐hedonically motivated (Grisanzio et al. 2021; Riediger et al. 2009). For instance, upregulating anger or fear may help prepare for confrontation or avoiding threats (Tamir et al. 2008; Tamir and Ford 2009), and negative stimuli may help satisfy morbid curiosity and cope with future stressors (Kerr et al. 2019; Oosterwijk 2017). Further, people may pursue short‐term negative emotions as a goal in itself (Inbar et al. 2013; Nelissen and Zeelenberg 2009). That is, contra‐hedonic motivations need not serve a functional goal: individuals may simply want to feel bad (Riediger et al. 2009), known as “benign masochism” (Rozin et al. 2013).

Contra‐hedonic motivations may help explain the shared developmental trajectory of ED behaviors and NSSI in adolescence, which is characterized by heightened experiences of negative emotions and a greater desire to maintain/increase negative affect and dampen positive affect (compared to adults) (Riediger et al. 2009). This may reflect adolescence being a period of emotional exploration (Riediger et al. 2009) and social bond formation, during which negative emotions may facilitate peer connection (e.g., via co‐rumination; Somerville and McLaughlin 2018). The concept of contra‐hedonic motivations is also aligned with work suggesting individuals engaging in NSSI are highly self‐critical, feel deserving of pain/punishment, and endorse self‐punishment motivations for self‐harm (Fox et al. 2017, 2019).

We aimed to (a) establish the presence of contra‐hedonic functions of ED behaviors (restrictive eating, binge eating, purging [including self‐induced vomiting, laxative use, driven/compulsive exercise]) and NSSI in a transdiagnostic sample of adolescents, (b) test for differences in endorsement of contra‐hedonic functions across behaviors, and (c) examine the relationship between contra‐hedonic and pro‐hedonic functions of these behaviors. We did not have a priori hypotheses.

2. Method

2.1. Participants and Procedure

Participants were recruited via a three‐wave longitudinal study (see Wang et al. 2021 for full details). Eligibility included: 12–14 years old at baseline, speaking English, living in the US, and reporting 2+ episodes of past‐month restrictive eating. Procedures were approved by the Committee on the use of Human Subjects at Harvard University. Items assessing contra‐hedonic functions were added to the survey midway through the longitudinal study. Thus, most participants completed these items in Wave 3 (with a few unique participants completing them in Wave 2), resulting in N = 137 participants for present analyses (see Table 1 for demographics).

TABLE 1.

Demographic and clinical characteristics information for the present sample.

Characteristic/behavior	Overall
	N = 137
Age (SD)	15.1 (0.86)
BMI (SD)	24.13 (7.01)
Race (%)
White	104 (76%)
Black	8 (6%)
Asian	15 (11%)
Native American/Pacific Islander	13 (9%)
Multi‐Racial	10 (7%)
Other	3 (2%)
Ethnicity
Hispanic	13 (9%)
Gender (%)
Cisgender female	103 (75%)
Cisgender male	5 (4%)
Transgender male	11 (8%)
Sexual orientation
Straight	32 (23%)
Gay/Lesbian	20 (15%)
Bisexual	51 (37%)
Not sure	12 (9%)
Not listed	22 (16%)
Behavior
Restrictive eating (%)	115 (83.94%)
Average episodes (SD)	27.08 (37.68)
Binge eating (%)	78 (56.93%)
Average episodes (SD)	8.25 (14.05)
Purging (%)	37 (27.01%)
Average episodes (SD)	3 (10.95)
NSSI (%)	70 (51.09%)
Average episodes (SD)	4.29 (10.48)
YEDE‐Q score (SD)	3.39 (1.26)

Note: In the present study, when we asked participants about restrictive eating, we defined this as, “any time you intentionally eat less than seems appropriate for the situation out of concern for your body shape and/or weight.” We also provided some examples (e.g., “eating an apple for dinner or fasting all day”). For binge eating, we defined this as “times when you have you eaten what other people would regard as an unusually large amount of food (given the circumstances) AND had a sense of lost control over your eating.” For purging, we defined this as “times when you have done any of the following as a means of controlling your shape or weight: made yourself sick (e.g., vomited), taken laxatives, exercised in a ‘driven’ or ‘compulsive’ way.” Average episodes (SD) refers to past 3 months frequency. Further, the sample's average YEDE‐Q score was above a recommended clinical cut‐off of 2.8 for adolescents (Velkoff et al. 2023).

2.2. Measures

Participants self‐reported age, race/ethnicity, gender, and sexual orientation. Participants completed the Youth Eating Disorders Examination‐Questionnaire (YEDE‐Q) (Goldschmidt et al. 2007), an abbreviated Self‐Injurious Thoughts and Behaviors Interview–Revised (SITBI‐R) (Fox et al. 2020), and the Self‐Rating Scale (SRS) to assess self‐criticism (Hooley et al. 2010) (including one additional item on the SRS to assess self‐punishment). We used an adapted Functional Assessment of Maladaptive Behaviors (FAMB) to assess functions of ED behaviors and NSSI (Wang et al. 2021), and added 8 items assessing contra‐hedonic functions (Table 2). Participants completed a FAMB for any behavior (restrictive eating, binge eating, purging, NSSI) endorsed over the past 3 months. Participants indicated how often they engaged in each behavior for each function, from “Never” (0) to “Often” (3).

TABLE 2.

Endorsement of contra‐hedonic items across behaviors.

Contrast	Estimate	SE	df	t ratio	p
To make yourself feel bad
Restrict—binge	0.10	0.13	201	0.73	0.89
Restrict—purge	−0.07	0.18	215	−0.37	0.98
Restrict—NSSI	−0.40	0.14	204	−2.92	0.02**
Binge—purge	−0.16	0.19	210	−0.87	0.82
Binge—NSSI	−0.50	0.15	212	−3.29	< 0.01**
Purge—NSSI	−0.34	0.19	210	−1.78	0.29
To make yourself feel angry, sad, lonely, anxious, etc.
Restrict—binge	0.13	0.12	199	1.13	0.67
Restrict—purge	0.15	0.16	213	0.94	0.78
Restrict—NSSI	−0.32	0.12	204	−2.63	0.04**
Binge—purge	0.02	0.16	208	0.09	1.00
Binge—NSSI	−0.45	0.14	211	−3.37	< 0.01**
Purge—NSSI	−0.47	0.17	207	−2.82	0.03**
To increase bad feelings
Restrict—binge	0.08	0.12	196	0.66	0.91
Restrict—purge	0.22	0.16	207	1.40	0.50
Restrict—NSSI	−0.33	0.12	200	−2.65	0.04**
Binge—purge	0.14	0.17	204	0.89	0.83
Binge—NSSI	−0.41	0.14	206	−2.98	0.02**
Purge—NSSI	−0.55	0.17	202	−3.25	< 0.01**
To increase feelings of anger, sadness, loneliness, anxiety, etc.
Restrict—binge	−0.03	0.11	192	−0.25	0.99
Restrict—purge	0.11	0.15	200	0.78	0.87
Restrict—NSSI	−0.43	0.12	196	−3.72	< 0.01**
Binge—purge	0.14	0.15	196	0.92	0.80
Binge—NSSI	−0.41	0.13	201	−3.17	< 0.01**
Purge—NSSI	−0.55	0.16	196	−3.46	< 0.01**
To keep feeling bad
Restrict—binge	0.21	0.13	204	1.58	0.39
Restrict—purge	0.30	0.18	220	1.70	0.33
Restrict—NSSI	−0.30	0.14	209	−2.19	0.13
Binge—purge	0.09	0.19	215	0.48	0.96
Binge—NSSI	−0.51	0.15	217	−3.36	< 0.01**
Purge—NSSI	−0.60	0.19	214	−3.18	< 0.01**
To feel less happy, good, or proud of yourself
Restrict—binge	−0.01	0.14	210	−0.10	1.00
Restrict—purge	0.06	0.19	234	0.31	0.99
Restrict—NSSI	−0.12	0.15	216	−0.76	0.87
Binge—purge	0.07	0.20	230	0.37	0.98
Binge—NSSI	−0.10	0.17	228	−0.61	0.93
Purge—NSSI	−0.17	0.21	229	−0.85	0.83
To stop feeling good
Restrict—binge	0.09	0.11	191	0.88	0.81
Restrict—purge	0.27	0.14	200	1.87	0.24
Restrict—NSSI	−0.05	0.11	196	−0.45	0.97
Binge—purge	0.17	0.15	196	1.15	0.67
Binge—NSSI	−0.14	0.13	202	−1.16	0.65
Purge—NSSI	−0.32	0.15	195	−2.07	0.17
To keep feeling angry, sad, lonely, anxious, etc.
Restrict—binge	0.30	0.13	200	2.35	0.09
Restrict—purge	0.38	0.17	214	2.24	0.12
Restrict—NSSI	−0.14	0.14	204	−1.04	0.72
Binge—purge	0.08	0.18	209	0.44	0.97
Binge—NSSI	−0.44	0.15	212	−2.99	0.02**
Purge—NSSI	−0.52	0.18	209	−2.85	0.02**

Note: Degrees‐of‐freedom method: Kenward‐roger. p value adjustment: Tukey method for comparing a family of 4 estimates. Estimates reflect mean differences in endorsement. A positive estimate indicates that the first behavior was more highly endorsed than the second; a negative estimate indicates that the second behavior was more highly endorsed than the first. A correspondent p‐value of less than 0.05 indicates that this difference was statistically significant after correction, and this is indicated by ** after those p‐values. All correlation matrices are in Figures 2, 3, 4, 5 and reflect correlations for individuals endorsing that behavior.

Abbreviations: df, degrees of freedom; NSSI, non‐suicidal self‐injury; SE, standard error.

2.3. Statistical Analysis

Analyses were performed in R (R Core Team 2021). First, we conducted one‐sample t‐tests evaluating whether mean scores of contra‐hedonic items were significantly greater than zero (“never”). Second, we conducted linear mixed effects models (LMMs) to explore whether endorsement of contra‐hedonic items differed across behaviors. Behavior type was entered as a categorical fixed effect using simple coding, with participant ID as a random effect (accounting for multiple behaviors nested within individuals). We conducted ANOVAs to test if the overall fixed effect of the behavior was significant, followed by pairwise comparisons of estimated marginal means. We corrected for multiple comparisons using Tukey's Honestly Significant Difference test. Finally, to explore the relationship between contra‐hedonic functions with pro‐hedonic functions and other clinical characteristics, we conducted correlational analyses at the item level, correcting for multiple comparisons using the Benjamini‐Hochberg procedure.

There were two types of missing data. First, some data were missing by design (e.g., participants who did not endorse binge eating were not shown the FAMB for binge eating, see Figures 2, 3, 4, 5). Second, there was item‐level missingness (from participants skipping items), which ranged from 2% to 4% missingness per item. For mixed effects models, observations with missing values were omitted prior to estimation, and models were estimated using restricted maximum likelihood estimation. Correlational analyses were conducted using pairwise deletion. Analysis code is available at https://github.com/francesghart/contrahedonicFunctions.

FIGURE 2.

Correlation matrix of pro‐hedonic functions (ANR = automatic negative reinforcement; APR = automatic positive reinforcement; SNR = social negative reinforcement; SPR = social positive reinforcement), contra‐hedonic functions, self‐punishment, self‐criticism, and frequency of behaviors for restrictive eating (n = 114). Values in each colored cell represent the Pearson r values of associations whose p values remained statistically significant after correction via Benjamini–Hochberg false discovery rate. Grayed out cells represent associations that did not reach statistical significance (i.e., p ≥ 0.05).

FIGURE 3.

Correlation matrix of pro‐hedonic functions (ANR = automatic negative reinforcement; APR = automatic positive reinforcement; SNR = social negative reinforcement; SPR = social positive reinforcement), contra‐hedonic functions, self‐punishment, self‐criticism, and frequency of behaviors for purging (n = 36). Values in each colored cell represent the Pearson r values of associations whose p values remained statistically significant after correction via Benjamini–Hochberg false discovery rate. Grayed out cells represent associations that did not reach statistical significance (i.e., p ≥ 0.05).

FIGURE 4.

Correlation matrix of pro‐hedonic functions (ANR = automatic negative reinforcement; APR = automatic positive reinforcement; SNR = social negative reinforcement; SPR = social positive reinforcement), contra‐hedonic functions, self‐punishment, self‐criticism, and frequency of behaviors for binge eating (n = 77). Values in each colored cell represent the Pearson r values of associations whose p values remained statistically significant after correction via Benjamini–Hochberg false discovery rate. Grayed out cells represent associations that did not reach statistical significance (i.e., p ≥ 0.05).

FIGURE 5.

Correlation matrix of pro‐hedonic functions (ANR = automatic negative reinforcement; APR = automatic positive reinforcement; SNR = social negative reinforcement; SPR = social positive reinforcement), contra‐hedonic functions, self‐punishment, self‐criticism, and frequency of behaviors for non‐suicidal self‐injury (NSSI; n = 67). Values in each colored cell represent the Pearson r values of associations whose p values remained statistically significant after correction via Benjamini–Hochberg false discovery rate. Grayed out cells represent associations that did not reach statistical significance (i.e., p ≥ 0.05).

3. Results

Participant demographics and clinical characteristics are in Table 1.

3.1. Endorsement of Contra‐Hedonic Functions

Nearly all participants (n = 111, 81%) endorsed at least one contra‐hedonic item. One‐sample t‐tests revealed endorsement of all contra‐hedonic items differed significantly from 0 (ts = 2.68–8.70, all ps ≤ 0.011) (Figure 1). The most highly endorsed items varied per behavior, with the top function for bingeing being “To feel less happy, good, or proud of yourself”; for NSSI, “To keep feeling bad”; and for purging and restricting, “To make yourself feel bad.”

FIGURE 1.

Stacked bar charts showing frequency of endorsement of contra‐hedonic functions of restrictive eating, binge eating, purging, and nonsuicidal self‐injury (NSSI).

3.2. Comparisons of Contra‐Hedonic Functions Across NSSI and ED Behaviors

LMMs revealed differences in endorsement of contra‐hedonic items across behaviors, with highest endorsement for NSSI compared to ED behaviors for nearly all items (Table 2). Specifically, the items “To make yourself feel angry, sad, lonely, anxious, etc.,” “To increase bad feelings,” and “To increase feelings of anger, sadness, loneliness, anxiety, etc.” had significantly higher endorsement for NSSI than all ED behaviors. The items “To keep feeling bad” and “To keep feeling angry, sad, lonely, anxious, etc.” also had significantly higher endorsement for NSSI than binge eating and purging. Finally, the item “To make yourself feel bad” had significantly higher endorsement for NSSI than restrictive eating and binge eating. No significant differences emerged across behaviors for other items.

3.3. Associations Between Contra‐Hedonic and Pro‐Hedonic Functions

For restrictive eating, nearly all contra‐hedonic items were significantly and positively correlated with pro‐hedonic social functions and automatic negative reinforcement. Most items were also correlated with desire to self‐punish via restrictive eating, self‐criticism, and restrictive eating frequency (Figures 2, 3, 4, 5).

For NSSI, nearly all contra‐hedonic items were significantly and positively correlated with pro‐hedonic social functions and automatic positive reinforcement, but not with automatic negative reinforcement. Most items were also correlated with desire to self‐punish via NSSI, but not self‐criticism or NSSI frequency.

For binge eating, nearly all contra‐hedonic items were significantly and positively correlated with pro‐hedonic social functions, with fewer significant correlations emerging for automatic positive and none for automatic negative reinforcement. Most items were also correlated with desire to self‐punish via binge eating, but not self‐criticism or binge eating frequency.

For purging, few significant associations emerged between contra‐hedonic items and pro‐hedonic functions or purging frequency. None were correlated with desire to self‐punish via purging or self‐criticism.

4. Discussion

This study presents new evidence for contra‐hedonic functions of restrictive eating, binge eating, purging, and NSSI among adolescents. Notably, we document that contra‐hedonic functions of ED behaviors and NSSI are not only present but prevalent. Most participants (> 80%) endorsed at least one contra‐hedonic motivation, and many (15%–40%) reported engaging in ED behaviors or NSSI to increase/maintain negative affect or decrease positive affect “sometimes” or “often.” While traditional emotion regulation models suggest people engage in self‐destructive behaviors to feel better, our results demonstrate people also engage in these behaviors to feel worse. Thus, a comprehensive framework of emotion regulation in psychopathology should account for both pro‐ and contra‐hedonic functions across individuals and contexts.

We also found considerable variability in contra‐hedonic functions across behaviors, with consistently stronger endorsement in contra‐hedonic functions for NSSI than ED behaviors. This may have important clinical implications, as the prominence of contra‐hedonic motivations in NSSI underscores the need for interventions that address the reinforcement of negative affect (e.g., exploring why individuals want to feel bad, and encouraging patients to utilize activities that achieve similar effects via non‐self‐destructive means).

Finally, we found that individuals engage in self‐destructive behaviors for pro‐ and contra‐hedonic motivations. Across behaviors, the more individuals that reported engaging in behaviors to reduce unwanted social interactions, the more that they also reported engaging in the same behavior to increase/maintain negative affect and reduce positive affect. While apparently paradoxical, one explanation is that perhaps individuals use the same behaviors at different times for different functions. For example, someone may restrict their eating at one time to feel good (e.g., for exerting self‐control) but at another to feel bad (e.g., believing they do not “deserve” to eat). Future research should use ecological momentary assessment (EMA) to evaluate the temporal dynamics of these functions.

These potential explanations notwithstanding, one may still wonder why anyone would want to feel worse. When considering emotions as instruments toward the pursuit of self‐directed goals (Tamir et al. 2008), perhaps the strongest explanation is in the relationship between contra‐hedonic functions and desire to self‐punish. Though historically assumed to be pro‐hedonic (e.g., feeling deserving of punishment and seeking relief via self‐inflicted pain), our results suggest people may use self‐punishment to harm themselves emotionally. Another candidate explanation is that adolescents may be motivated to maintain/worsen emotional states to communicate that their suffering is “valid.” This is aligned with work demonstrating that individuals with self‐destructive behaviors often struggle with emotional validation and not feeling “sick enough” (Eiring et al. 2021). Thus, some individuals may want to feel worse to validate that their emotional pain is real, severe, and worthy of attention—complementing existing evidence that negative emotions can strengthen interpersonal relationships and signal group membership (Brady and Van Bavel 2025; Somerville and McLaughlin 2018). This effect may be particularly powerful in adolescence, during which group‐ and identity‐related challenges are salient.

Our findings should be interpreted considering several limitations. First, the sample size was relatively small, particularly for purging, limiting statistical power. This was partially due to our data being drawn from a larger study focused on restrictive eating. Additionally, given that people engaging in disordered eating often report automatic negative reinforcement (including self‐punishment) motivations for driven exercise (Kolar et al. 2025), future work should explore potential relationships between these motivations and contra‐hedonic motivations in a larger sample of individuals endorsing purging behaviors. Second, the sample was predominantly White, restricting generalizability. Third, the retrospective self‐report introduces potential recall bias. Future research should leverage EMA to capture real‐time fluctuations in contra‐hedonic motivations. Fourth, the cross‐sectional data preclude causal inferences. Developmental factors also warrant further investigation, as contra‐hedonic motivations for self‐harming behaviors in adulthood remain unclear. Finally, we intentionally used item‐level analysis (given the exploratory aim to probe the existence of contra‐hedonic motivations). While aligned with best‐practice recommendations in measurement to explore a construct's existence, breadth, and depth before developing and validating measures (Fried et al. 2022; Wang et al. 2025), future research should develop a valid and reliable self‐report measure of contra‐hedonic functions of self‐destructive behaviors.

This study reveals a striking and underrecognized dimension of emotion regulation: some adolescents engage in ED behaviors and NSSI not to escape negative emotions, but to amplify them. These findings challenge prevailing assumptions that self‐destructive behaviors function to reduce aversive states and suggest a more complex psychological function: using self‐harm to intensify emotional pain. We believe theoretical models of ED behaviors and NSSI should expand to incorporate this contra‐hedonic dimension. Clinically, interventions may miss a critical target if they assume behaviors only ever serve to reduce emotional pain. Future research should investigate how contra‐hedonic motives unfold over time in people's daily lives, and how interventions might disrupt these cycles of self‐directed harm.

Author Contributions

Frances Grace Hart: formal analysis, writing – original draft, writing – review and editing, visualization, project administration. Sharina Hamm: formal analysis, writing – review and editing. Leily M. Behbehani: visualization, writing – review and editing. Kathryn R. Fox: conceptualization, methodology, writing – review and editing, project administration, data curation. Kathryn A. Coniglio: conceptualization, methodology, writing – review and editing. Erin E. Reilly: conceptualization, methodology, writing – review and editing. Leah H. Somerville: conceptualization, methodology, writing – review and editing. Shirley B. Wang: conceptualization, methodology, formal analysis, writing – original draft, writing – review and editing, data curation, project administration, supervision.

Lived Experience Involvement Statement

No specific efforts were undertaken to involve persons with lived experience in the study design or execution, or in the preparation of this manuscript.

Funding

This work was supported by the National Institutes of Health (K23MH131871); UCSF Weill Neurohub; UC Society of Hellman Fellows; UCSF Research Evaluation and Allocation Committee (REAC), John E. Cox Fund.

Disclosure

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Hart, F. G. , Hamm S., Behbehani L. M., et al. 2026. “Contra‐Hedonic Functions of Self‐Destructive Behaviors in Adolescents.” International Journal of Eating Disorders 59, no. 8: 1776–1784. 10.1002/eat.70111.

Data Availability Statement

Data and materials are available upon request. Analytic code is available at the first author's GitHub (https://github.com/francesghart/contrahedonicFunctions).