Abstract
Introduction
The relationship between disordered eating (DE) and shame is established, including body shame, shame around eating, internal shame, binge eating-related shame, external shame and body image shame (BIS) (Nechita et al., 2021; O’Loghlen et al., 2022). Further, a 2016 systematic review of 28 studies found a positive association between DE and both external shame and BIS, and a negative association between DE and self-compassion (SC) (Braun et al., 2016).
Shame is an unpleasant emotion that involves feelings of unworthiness and beliefs that others are judging one negatively (Braun et al., 2016; O’Loghlen et al., 2022; Skarderud, 2007). Internal shame involves an inwardly focussed sense of inadequacy and inferiority compared to others, internal BIS involves feelings of inadequacy and embarrassment regarding body shape and weight, binge eating-related shame involves feelings of shame and disgust about one's binge eating behaviour, external shame involves the perception that others see one as flawed or inferior, and external body shame involves the belief that one is evaluated negatively because of body shape and weight (O’Loghlen et al., 2022). Shame contrasts with SC, which involves self-kindness, acknowledging pain as part of the common human experience, and not defining oneself by painful thoughts (Neff, 2003a). Shame is also different to guilt, even though the terms are often used interchangeably. Shame can be defined as an intense painful feeling or experience that leads us to believe that we are flawed and unworthy, while guilt might be an adaptive and even helpful emotion that we feel when we have done something out of alignment with our values (Brown, 2013).
So far, research that integrates DE, shame and SC has only involved narrow participant demographics. For example, men have been poorly represented, as have populations outside universities and non-Western countries (Braun et al., 2016).
This exploratory sequential mixed methods cross-sectional study aimed to fill the above research gaps by investigating associations between – and perceptions of – DE, external and BIS, and SC; via an online survey in an international, general sample of both men and women. We hypothesised that DE would be positively correlated with external shame and BIS, and negatively correlated with SC (Braun et al., 2016); and that there may or may not be sex differences in outcome variables (Nechita et al., 2021).
Methods
Procedure and participants
This study was approved by x, reference number x. It was an exploratory mixed methods study which was sequential-independent and quantitatively driven (Schoonenboom and Johnson, 2017). There were two phases of the study, quantitative Phase 1 (P1) and qualitative Phase 2 (P2). Each phase involved a separate online survey. The phases were linked in that some participants from P1 were given the option of completing P2, as is described in more detail below, and the topics explored in P1 informed P2's four open-ended survey questions. At the results stage, P1 and P2 data were integrated.
The online surveys were hosted on Qualtrics.com. Participants were recruited through two separate posts on Amazon Mechanical Turk (MTurk), an online platform enabling sampling of an international population. All participants gave informed consent by checking a box alongside the words ‘I agree, start survey’.
Inclusion criteria for our study were aged 18 years + and English as first language. Participants who completed P1 of the survey were automatically reimbursed for their time with $1USD by MTurk, even if their responses were later removed during data cleaning. Participants who adequately completed P2 of the survey were reimbursed $5USD (first 20 participants) or $1USD (the remaining participants) through MTurk, after their answers were checked for coherence. The reimbursement amounts were based on the US hourly minimum wage and the time estimated to complete each part of the survey.
P1
We collected the following data during May and June in 2018: age, sex and location. We also assessed levels of DE (via the Eating Attitudes Test-26 (EAT-26) (Garner et al., 1982)), external (Other as Shamer-2 (OAS-2) (Matos et al., 2015) and BIS (Body Image Shame Scale (BISS) (Duarte et al., 2015), and SC (Self-compassion Scale – Short Form (SCS-SF) which has six sub-scales: Self-Kindness, Self-Judgement, Common Humanity, Isolation, Mindfulness, and Overidentification (Costa et al., 2016; Raes et al., 2011)) – as outlined in the Supplemental material. All scales have previously displayed good internal reliability and construct validity (Duarte et al., 2015; Garner et al., 1982; Matos et al., 2015; Raes et al., 2011). The internal consistency reliability coefficients of the scales in the current study were high: EAT-26 α = 0.90, OAS-2 α = 0.92, BISS α = 0.96, and SCS-SF α = 0.86.
Qualtrics additionally provided survey metadata, for example, time taken to complete.
Statistical aspects
Details on how our sample size was calculated and data cleaning are provided in the Supplemental material.
Our primary outcome was levels of external shame using the OAS-2 (Matos et al., 2015), and we predicted this would be significantly higher in participants with EAT-26 scores >20 compared to those with EAT-26 scores 20 or less (EAT-26, 2021; Garner et al., 1982). Based on previous research, we additionally hypothesised that DE would be positively correlated with external shame and BIS, and negatively correlated with SC (Braun et al., 2016), and that there may or may not be sex differences in outcome variables (Nechita et al., 2021).
Descriptive statistics were calculated. Skewness and kurtosis were calculated for all continuous scales and none showed significant deviation from normal distribution (skewness < |3|, kurtosis < |10|) (Kline, 2010).
Independent sample t-tests were used to compare the main outcome scale variables (EAT-26, OAS-2, BISS, and SCS-SF) between: EAT-26-derived level of concern risk of developing an ED or not, and sexes. When comparing outcomes between sexes, we removed the two participants who selected ‘other’ sex from analysis due to small sample size.
Pearson correlation coefficients were calculated to assess the relationships between the four scales, as well as between SCS-SF and age. The strength of the relationships was based on psychology terms as per Dancey and Reidy (2007). We also ran partial correlations for total scores of the following, while adjusting for sex and age: EAT-26 and OAS-2, EAT-26 and BISS, and EAT-26 and SCS-SF.
Results were considered significant if p < 0.05.
Linking the phases
Of the participants who provided reliable answers to the P1 survey and who expressed an interested in completing the P2 survey (67/142), 23 had EAT-26 scores of 20 or more and were emailed information about P2. This was because we wanted to ensure that we included some people with DE behaviours in P2 because our online recruitment targeted the general population (which would also comprise people with DE behaviours).
Additionally, the topics explored in P1 informed the questions that were asked in P2.
P2
Responses to this part of the survey were collected during 19–29 August 2018. We first asked for the following participant information: age, location, name, and email. Qualtrics additionally provided survey metadata. We then asked four free-response questions (Qs), and each required a minimum answer of 200 characters: (Q1) In the modern world, there are people who struggle with DE. What do you think about this issue? (Q2) What does ‘shame’ and ‘SC’ mean to you, and how do you think they interact with your eating habits? (Q3) What are some strategies that you employ to reduce feelings of shame and become more self-compassionate? Please think about these factors in relation to your eating. (Q4) When you feel you need extra help from others, who are some people you would talk to? What may prevent you from talking with those who can help you?
These questions intended to explore the psychological topics from P1 in greater depth to explore: how the general population perceives shame, SC and DE; whether the general population is aware of the links between these factors; and informing potential interventions. We initially aimed for a sample size of 20, but increased this to 100 once we had received 20 completed surveys, because the respondents were not writing as much as we had expected; and because many responses were non-sensical copy-and-paste answers that had been Googled.
Author ZYW, a female undergraduate medical student at the time, conducted the qualitative analysis as part of her Honours research. She had no personal experience of DE and focused her analysis on identifying any links between DE, shame and SC. Inductive thematic analysis was carried out according to guidelines provided by Braun and Clarke (2006). The free text data were imported into NVivo software and were read to ensure familiarity. Next, codes were generated from repeating ideas and patterns, and were refined through re-reading and re-coding. These refined codes were checked with an external qualitative expert (AG), who confirmed that the coding was appropriate, and preliminary themes were developed. The themes were derived from the answers only, and not aligned with pre-determined theoretical frameworks. The themes were revised to ensure they were representative of the data.
Results
P1
There were 219 participants who started P1 of the online survey, with one not providing consent, and 13 not meeting the inclusion criteria; leaving 205 completers. Of these, 63 were omitted during data cleaning for providing unusual response patterns, which were assessed as being unreliable, leaving 142 for analysis.
When we compared duration of survey completion between the 142 included responses and the 63 excluded unusual response patterns, the excluded responses were completed significantly faster (mean 360 ± SD 198 s) than the included responses (mean 454 ± SD 237 s) (t = −2.737, p < 0.01). This suggested that omitted answers were more likely to be completed without reading the questions properly in order to rush through the questionnaire so as to be paid for study participation.
Demographics
Of the 142 participants, 82 (57.7%) were male, 58 (40.8%) were female, and 2 (1.4%) identified as ‘other’ sex. Mean age was 33 ± SD 10 years, range 21–64 years. Most participants were from the US (96/142, 67.6%), followed by India (43/142, 30.3%), and Canada (3/142, 2.1%).
Classifying participants into high versus low level of concern about dieting, body weight or eating behaviour based on EAT-26 scores
See Table 1.
Risk of an ED versus external shame, BIS and SC.
BIS: body image shame; BISS: BIS Scale (Duarte et al., 2015); ED: eating disorder; OAS-2: Other-as-Shamer-2 scale (Matos et al., 2015); SCS-SF: SC Scale – Short Form (Costa et al., 2016); SD: standard deviation.
High risk of an ED is defined as an Eating Attitudes Test-26 (EAT-26) score of 20+, classified as not high risk of an ED with an EAT-26 score of <20 (Garner et al., 1982).
Equal variances not assumed.
** = p < 0.01, *** = p < 0.001.
Associations between disordered eating, shame and self-compassion
See Table 2.
Mean scale scores between men and women, and associations between DE, external shame, BIS, and SC in all participants.
BIS: body image shame; BISS: BIS Scale (Duarte et al., 2015); DE: disordered eating; EAT-26 = Eating Attitudes Test-26 (Garner et al., 1982); OAS-2: Other-as-Shamer-2 scale (Matos et al., 2015); SCS-SF: SC Scale – Short Form (Costa et al., 2016); SD: standard deviation.
Partial correlation adjusted for sex.
Partial correlation adjusted for age.
* = p < 0.05, *** = p < 0.001.
Sex differences in study variables
See Table 2.
Correlation of age with self-compassion
A Pearson correlation between age and SC showed a non-significant (p = 0.339) lack of association between these variables (r = 0.08).
P2
Ninety-two participants completed P2 of our online survey, with nine excluded because they did not meet the inclusion criteria, and 28 responses rejected because they were incoherent or irrelevant. An example of part of a rejected response to Q1: ‘People with anorexia and bulimia often have a strange desire to be near the… Kitchen, recently admitted that she struggled with an ED… You might also like’.
Therefore, 55 responses were included in analyses. These participants were aged 20–68 years, with a mean age of 35 ± SD 11 years. Nearly all participants lived in the US (51, 93%), three in India (5%), and one in Switzerland (2%).
Four key themes (T) were identified after conducting inductive thematic analysis on all four questions together: (T1) DE as a Psychological Issue, (T2) DE as a Social Issue, (T3) The Vicious Cycle of Shame and (T4) Seeking Help (Supplemental material). These themes do not directly correspond with the four survey questions.
We were unable to extract substantial information regarding SC from the free-text responses. This is despite participants being asked what SC means to them in (Q2). Many participants expressed that they ‘have never heard of this term’, and acknowledged that they can only guess it to mean ‘being kind to yourself’, which was insufficient information to constitute a theme. This is consistent with Neff (2003b) identifying that SC originated from Eastern Buddhist philosophies and is not well-integrated into Western culture – where most of our P2 participants were from (the US).
Discussion and integrating the study phases
The current mixed methods cross-sectional study investigated associations between, and perceptions of, DE, external and BIS, and SC in an community sample of younger men and women who were mainly residing in the US and India. Amongst the 142 participants that we surveyed in the quantitative P1 of our study, we reported in multiple ways that DE was significantly correlated with more shame, both external and body image-specific. These results are backed up by findings from the qualitative P2 of our study, where 55 younger male and female participants, also mainly residing in the US, highlighted detrimental societal and personal connections between DE and shame.
Our results support our primary outcome hypothesis of OAS-2 external shame scores being significantly higher in participants who had EAT-26 scores of >20 versus those who had EAT-26 scores of <20 (Garner et al., 1982). The people displaying worrying dieting, body weight, and eating behaviours with EAT-26 scores of >20 also reported significantly higher levels of BIS, and self-compassion-related self-judgement and overidentification sub-scales (please read below for elaboration upon the SC sub-scales). The shame findings were backed up by correlational results from our study of higher EAT-26 scores being significantly and moderately associated with higher OAS-2 and BISS scores. When controlling for sex and age in partial correlation analyses, the moderate and significant associations between DE and both types of shame remained, indicating that sex and age had little influence on the relationship between DE and both types of shame. Additionally, both forms of shame were significantly correlated, i.e. more external shame was associated with more body image-specific shame. That is, within the shame-DE cycle, DE may increase external shame and BIS, and vice versa. These findings are consistent with our qualitative results, where participants highlight the psychological and social drivers behind DE, with a key theme of shame in the complex web of risk.
Focusing further on shame, our findings are also consistent with the meta-analysis by Nechita et al. (2021) linking shame and ED symptoms, and the 2021 review by O’Loghlen et al. linking shame and binge eating symptoms specifically. Nechita et al. (2021) found that ED symptoms were significantly related to shame, especially body shame, with an average medium effect size (using Pearson correlation coefficient) of 0.55, while we report a correlation between EAT-26 score and BISS score of 0.52. If we look to specific examples of individual studies, such as Ferreria et al.'s (2013) study involving women from the general population in Portugal, and the Mustapic et al. (2015) study in female secondary school students in Croatia, we report comparable moderate associations between DE behaviours, BIS (as measured by the body shame sub-scale of the objectified body consciousness scale, Mustapic et al., 2015), and OAS-measured external shame (Ferreria et al., 2013). Our shame results also align with the O’Loghlen et al. (2022) review of binge eating studies and shame, specifically that BIS and external shame are associated with binge eating symptoms. Higher EAT-26 scores in P1 of our study were associated with higher levels of both types of shame, and the EAT-26 questionnaire involves questions relating to binge eating, such as, ‘I have gone on eating binges where I feel that I may not be able to stop’. Participants in P2 of our study mentioned binge eating in themes (T1) and (T3) as an effect of negative emotions – like shame – and a cause of them. For example, ‘Food can help numb the pain’ (46 years, US).
Our DE-shame associations are further exemplified in participant comments such as, ‘…look in the mirror and … see ugliness’ (36 years, US) when talking about body-related shame, and, ‘I didn’t want to be judged harshly for my actions’ (39 years, US) when talking about not seeking help because of the fear of external shame.
Focusing on SC, our findings are in part as per the 2016 review by Braun et al. linking DE with less SC, with our finding of a significant, weak negative Pearson correlation between EAT-26 score and SCS-SF score of −0.17 – which is the same value that was also found between EAT-26 and SCS-SF by Taylor et al. (2015) in US college students. However, during partial correlation analyses with DE and SC, the significance of the weak, negative relationship was lost when controlling for both sex and age, suggesting that sex and age affect the relationship between DE and SC. We also reported that SC is significantly higher in men compared to women (Table 2), but that there was no significant correlation between SC and age. Hence, in our sample, it seems that sex affects SC more so than age. Additionally we found that people with EAT-26 scores of >20 also have significantly higher levels of SC-related self-judgement and overidentification. These two sub-scales of the SCS-SF comprise the following questions: Self-Judgement, ‘I’m disapproving and judgemental about my own flaws and inadequacies’ and ‘I’m intolerant and impatient towards those aspects of my personality I don’t like’; Overidentification, ‘When I fail at something important to me I become consumed by feelings of inadequacy’ and ‘When I’m feeling down I tend to obsess and fixate on everything that's wrong’ (Raes et al., 2011). We also reported that both external shame and BISS were significantly negatively correlated with SC. These findings corroborate previous results, such as the 2016 review by Braun et al. that highlighted SC as a protective factor in the DE-shame cycle. However, we did not find that SC was a significant factor in the DE-shame cycle in all aspects, for example, we did not find significant associations between DE and total SC during partial correlations that controlled for age and sex. Perhaps this can be explained by the finding from our qualitative investigation that participants did not understand the concept or application of SC. Also, perhaps if we had assessed compassion specific towards one's body, we may have found stronger correlations to DE, as demonstrated by Oliveira et al. (2018), who reported a significant, moderate inverse correlation of r = −0.47 between DE and body compassion as measured by the Body Compassion Scale.
Focusing more on sex differences, when we separated participants by sex, we found that men reported healthier levels on the measured study variables than women, that is, lower levels of eating pathology, less external shame and BIS, and more SC. This is consistent with some studies, such as Melo et al. (2020), where Portuguese women aged 18–40 years from the general population reported higher levels of binge eating, external and internal shame, and body shame, than men from the sample population. This is in contrast to the Nechita et al. (2021) meta-analysis that reported that the percentage of female and male participants in the analysis did not affect the shame-DE association, but that may have been because most of the studies were conducted on predominantly women with small samples of men.
Regarding the main coping strategy for any DE experienced by our sample, help-seeking, most participants found this action difficult – often because of shame. Until someone is ready to accept that they have a problem with DE, anonymous help via online means may be preferred. Once someone acknowledges that DE is a problem, opening up to someone who knows them and who they trust – friend, family, or even health professional – may be the next step. The listening and helping skills of the helper are also important, for example, can someone truly listen without judgment? And can anyone truly help DE of clinical severity if not a health professional? It was interesting that not many participants listed medical professionals as sources of support, and we do not know whether this is because of unhelpful previous experiences in a professional space, or lack of knowledge of professional help being available.
Future investigative studies should involve larger sample sizes that comprise half men to reflect a general population, across a wide variety of countries, addressing the limitations outlined below. Future intervention studies might focus on reducing external and body image-specific shame, and the Self-Judgement and Overidentification aspects of SC, to reduce DE behaviours in general and clinical populations of men and women. There may be utility in developing separate interventions for men and women if men are shown to have healthier baseline levels of eating behaviour, shame, and SC. Also, based on our qualitative P2 findings, interventions that aim to improve shame may also increase the likelihood of someone suffering from DE to seek help from others.
Strengths and limitations
Strengths of our novel study include: the large percentage of men in our study sample, an under-represented population in eating-shame research (Braun et al., 2016; Nechita et al., 2021); the recruitment of people from the general community, while most non-clinical research has recruited university students (Braun et al., 2016; Nechita et al., 2021); the involvement of participants from non-Western regions, such as India (where 30.3% of Study 1 participants were from), when most research has been conducted in Western regions (Braun et al., 2016; Nechita et al., 2021); the mixed methods design of our study which aimed to expand and strengthen our study's conclusions (Schoonenboom and Johnson, 2017); using MTurk as a reliable means of recruitment, for example, Buhrmester et al. (2011) reported that 116 MTurk participants demonstrated similar test-retest reliability coefficients on measures of self-esteem compared with results from a traditional data collection method in undergraduate students by Bosson et al. (2000) (test–retest reliability correlation 0.87; p < 0.01).
Limitations of our study include the cross-sectional nature of our study, that cannot show causation, nor does the study design permit the investigation of changes in factors over time; not asking for the gender of participants in P1, and neither gender nor sex in P2, which limited our discussion of sex-specific patterns between variables; there may have been sampling bias in our study, where individuals with experience of DE may have been more likely to participate in our study; there is bias inherent in the subjective self-report nature of surveys; the financial incentives for participants may have contributed to the high numbers of unreliable responses, with people possibly lying about age and first language being English, and not reading questions properly in order to be paid quickly for survey completion (this can be seen by the finding that survey responses that were excluded took less time to complete than responses that were included); the lack of validity and attention check questions in the survey to improve the reliability of the MTurk data; the fact that we only received responses from people living in four countries, mostly the US, possibly relating to the time that we published the surveys in Australia (2pm AEST) and the corresponding time in the included countries (10pm in US and Canada, and 8:30am in India); the fact that only 0.02% of the Indian population have English as their first language (Office of the Registrar General & Census Commissioner, 2011) suggests that our Indian sample was not representative of the Indian population; the inability to differentiate between personal experience versus opinion on study themes in P2; the limitations of binary categorisation of participants into EAT-26 risk categories; and the lack of correction for multiple comparisons between variables across categories.
Conclusion
The findings from the current study strengthen the evidence base on DE, shame, and SC by involving male participants recruited from the general community, including in India. In our participants, DE behaviours (such as binge eating) were related to more external shame (such as thinking that others don’t see you as good enough) and body image-specific shame (such as being worried about being criticised in public because of your body shape), and less SC. Men had healthier levels of study variables than women. The vicious cycle of pathological eating and shame might involve someone eating for emotional reasons, including binge eating, which can result in the unpleasant emotion of shame. Someone may then try to avoid this negative feeling by numbing themselves with more food.
Supplemental Material
sj-docx-1-nah-10.1177_02601060231201511 - Supplemental material for ‘Food can help numb the pain’: A mixed methods exploration of disordered eating, shame and self-compassion in an international community sample of men and women
Supplemental material, sj-docx-1-nah-10.1177_02601060231201511 for ‘Food can help numb the pain’: A mixed methods exploration of disordered eating, shame and self-compassion in an international community sample of men and women by Zhi Yi Wang, Patrick Rawstorne and Rebecca C Reynolds in Nutrition and Health
Supplemental Material
sj-pdf-2-nah-10.1177_02601060231201511 - Supplemental material for ‘Food can help numb the pain’: A mixed methods exploration of disordered eating, shame and self-compassion in an international community sample of men and women
Supplemental material, sj-pdf-2-nah-10.1177_02601060231201511 for ‘Food can help numb the pain’: A mixed methods exploration of disordered eating, shame and self-compassion in an international community sample of men and women by Zhi Yi Wang, Patrick Rawstorne and Rebecca C Reynolds in Nutrition and Health
Supplemental Material
sj-pdf-3-nah-10.1177_02601060231201511 - Supplemental material for ‘Food can help numb the pain’: A mixed methods exploration of disordered eating, shame and self-compassion in an international community sample of men and women
Supplemental material, sj-pdf-3-nah-10.1177_02601060231201511 for ‘Food can help numb the pain’: A mixed methods exploration of disordered eating, shame and self-compassion in an international community sample of men and women by Zhi Yi Wang, Patrick Rawstorne and Rebecca C Reynolds in Nutrition and Health
Footnotes
Acknowledgements
The authors would like to thank Professor Boaz Shulruf for his advice on quantitative analysis and Dr Alexandra Gibson (AG) for her advice on qualitative design and analysis.
Authors’ contributions
This research was for ZYW's undergraduate medicine Honours research project – for which RCR and PR provided supervision. RCR conceptualised and partly funded the research. ZYW and RCR collected the data and managed the project. All authors contributed to data analysis, with ZYW doing the majority of it. Amazon provided the MTurk online platform and UNSW Sydney provided access to Qualtrics, SPSS and NVivo software. ZYW and RCR wrote the original draft of the manuscript, and all authors reviewed and edited it.
Availability of data and materials
The raw data generated during the study is available via the following Figshare link: 10.6084/m9.figshare.24087072.
Consent for publication
All participants gave informed consent by affirming that they had read the participant information sheet, etc., by checking a box alongside the words ‘I agree, start survey’.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Ethical statement
This study was approved by the UNSW Human Research Ethics Committee, reference number HC18025.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: RCR received an internal Career Advancement Fund from UNSW Sydney in 2017, and used some of these funds to pay for participant reimbursement on MTurk. The authors received no other internal or external financial support for the research, authorship, and/or publication of this article from the public, commercial, or not-for-profit sectors.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
