Starving off death: Mortality salience impacts women’s body image and disordered eating

Abstract

From a Terror Management perspective, the pursuit of thinness that characterizes eating disorders (EDs) is arguably a culturally endorsed way to mitigate death anxiety. In the present studies, we used the mortality salience (MS) paradigm to examine whether priming death increases ED symptoms. We recruited two samples of women from undergraduate (Study 1, N = 120), and clinically relevant (Study 2, N = 154) populations. After priming, participants completed measures of potential confounds (e.g., neuroticism, affect). Next, we assessed ED attitudes and behavior using a portion size estimation task, and measures of body dissatisfaction and eating intention. Study 1 findings were inconsistent with the claimed role of death anxiety in ED related behavior. However, in Study 2, MS priming led to increased dissatisfaction with current thinness and smaller portion sizes for high-fat compared to low-fat food. Overall, the results suggest that death anxiety may, at least partially, drive ED symptoms.

Fears of death are a central part of the human condition, and have been considered “the worm at the core” of our existence (James, 1985, p. 119). Existential theorists have long suggested that difficulty in effectively buffering death anxiety may lead to the “extreme modes of defense” observed in psychological disorders (Yalom, 1980, p. 111). However, empirical investigation into the role of death anxiety in driving mental illness was only recently galvanized by Iverach et al. (2014)’s argument that death anxiety is a transdiagnostic construct. The authors propose that death anxiety underlies the development of numerous psychological disorders and might account for the “revolving door” of mental health (p. 590). If this is the case, then despite successful disorder-specific treatments, failing to address death anxiety may contribute to ongoing mental health issues (Iverach et al., 2014).

Cross-sectional research has demonstrated that death anxiety predicts broad markers of psychopathology (e.g., number of hospitalizations and medications, and lifetime number of disorders), and has large relationships with the symptom severity of 12 psychological disorders (Menzies et al., 2019). Experimental studies further support these cross-sectional relationships. Such studies have used the MS paradigm, in which participants are randomly allocated to a death prime or a control prime (e.g., dental pain; Burke et al., 2010) to measure the impact of death anxiety on disorder-consistent behavior. For example, among students with spider phobia, high levels of social anxiety, and contamination concerns, reminders of death increased anxious behaviors (e.g., avoidance; Strachan et al., 2007). Further, MS priming increased handwashing behavior among participants with the contamination subtype of obsessive compulsive disorder (OCD; Menzies & Dar-Nimrod, 2017), and health anxious behavior among those with anxiety-related disorders involving body-scanning (Menzies et al., 2021). Taken together, these results suggest that people employ specific, disorder-consistent defenses to protect against death anxiety.

The importance of identifying transdiagnostic constructs, such as death anxiety, has been increasingly recognized. In particular, eating disorders (EDs) is one diagnostic group in which transdiagnostic constructs have garnered much attention (e.g., perfectionism, Egan et al., 2011). Given the high mortality rate of EDs (Fichter & Quadflieg, 2016), and their significant comorbidity with other disorders and resistance to treatment (Udo & Grilo, 2019; van den Berg et al., 2019), further research focused on identifying transdiagnostic constructs in these disorders is sorely needed.

Death anxiety is one transdiagnostic process that could be related to the pursuit of thinness that characterizes many body image issues and EDs. Although the connection between fears of death and EDs may, at first, appear less obvious than in some disorders (e.g., OCD [Menzies et al., 2015] and illness anxiety disorder [Furer et al., 2007]), death anxiety has been shown to have equally strong relationships with disorders where there is no overt preoccupation with physical harm, such as depression and social anxiety (Menzies et al., 2019).

The somewhat esoteric relationship between death anxiety and EDs may be explained by Terror Management Theory (Greenberg et al., 1986). Numerous studies have demonstrated that humans manage fears of death through a dual-component cultural anxiety buffer (Pyszczynski et al., 1999). The components of this buffer include: (1) Cultural worldviews (i.e., shared belief systems or values), and (2) Self-esteem. Together, they enable individuals to view themselves persons of value who are symbolically transcending death by living up to their cultural standards (Pyszczynski et al., 1999).

The mortality salience (MS) hypothesis states that reminders of death will increase behaviors that fulfill one’s cultural worldview and bolster self-esteem, because these behaviors reassure people that their existence is meaningful (Arndt et al., 2005; Pyszczynski et al., 1999). Hundreds of TMT studies have tested this hypothesis and a meta-analysis of 277 studies found that in 80% of studies reminders of death increased behaviors that maintained cultural worldviews and self-esteem (Burke et al., 2010).

Consistent with TMT, thinness is a cultural standard in western societies that women, in particular, might strive for to bolster self-esteem and mitigate death anxiety. Indeed, internalizing the thin ideal is a significant risk factor for EDs among women (Argyrides et al., 2020). Theoretically, disordered eating may be one way that women can defend against fears of death, by aligning with a prevailing cultural worldview emphasizing the importance of thinness (Goldenberg et al., 2005; Iverach et al., 2014; Le Marne & Harris, 2016).

It may appear counterintuitive that a disorder like anorexia nervosa, with such high mortality rates (Fichter & Quadflieg, 2016), could be driven by the fear of death. However, decades of research have shown that reminders of death can increase life-threatening behavior if that behavior is important to an individual’s self-image. For example, MS primes increased risky driving behavior (Ben-Ari et al., 2000), sun-tanning (Routledge et al., 2004), and smoking (Hansen et al., 2010), but only for individuals who indicated that these behaviors are relevant to their self-esteem. Therefore, although disordered eating may appear contrary to prolonging one’s life, these behaviors may enable women to meet the standards promoted by their culture and bolster their self-esteem, thereby obtaining a sense of symbolic immortality in the face of death.

Few studies have explored the relationship between death anxiety and body image and ED related behavior. Research has shown that a preoccupation with death features prominently in the illness narratives of women with EDs (Farber et al., 2007) and that death anxiety is more severe among women diagnosed with anorexia nervosa than matched controls (Giles, 1996). Moreover, many individuals with EDs present with death-related ideation (Farber et al., 2007). Indeed, death anxiety is a stronger predictor of ED symptomatology than perfectionism, and remained an independent predictor after controlling for self-esteem, age and perfectionism (Le Marne & Harris, 2016). In one of the only experimental studies to date, a MS prime led undergraduate women, but not men, to eat less of a high-fat snack, and perceive themselves as further from their ideal thinness when the group setting was salient (Goldenberg et al., 2005). Contrary to these results are findings that death anxiety did not appear to drive the body dissatisfaction or altered eating behavior typical of muscle dysmorphia among men interested in health and fitness (Menzies et al., 2023). However, the results for muscle dysmorphia were qualified by post-hoc analyses demonstrating an increase in portion size after MS priming among those with high, but not low, muscle dysmorphia symptoms. Overall, these results suggest that death anxiety could play a role in EDs and body image concerns, but to date, the existing literature is sparse.

Although Goldenberg et al. (2005) study demonstrated a MS priming effect on eating behavior and body image, their research has some limitations. First, the effect of the MS prime was only examined on high-fat snacks. It remains unclear whether death anxiety drives reduced consumption overall or only for high-fat food. Second, researchers measured only ratings of perceived thinness, omitting other body shape related attitudes (e.g., muscularity). Third and fourth, this research involved only an undergraduate sample and did not assess ED symptomatology.

No previous research has investigated whether symptom severity moderates the effect of MS priming on ED attitudes and behavior. We expected that MS priming would have a larger effect on people with more severe ED symptoms. That is, when confronted with a reminder of death, people with more severe ED symptoms, would be more likely to employ disorder-consistent defenses to manage their anxiety. We therefore conducted two studies to address this gap in the literature. Study 1 employed undergraduate women. Study 2 used the same methods, employing a clinically relevant sample of women with body image concerns or disordered eating behavior. We predicted that participants in the MS priming condition would show greater body dissatisfaction and ED behavior compared to those in the control condition. We also predicted that participants with more severe ED symptoms would show a greater impact of the MS prime than those with less severe ED symptoms.

Overview of the Overall Study designs

We recruited two samples. The sample for Study 1 comprised undergraduate women, and the sample for Study 2 specifically included women who identified as having clinically relevant concerns about their body image or eating behavior. Being female was an inclusion criterion because the thin ideal is more salient among women than men (Goldenberg et al., 2005; Le Marne & Harris, 2016). We determined the sample size through an a priori power analysis using G*Power software (Faul et al., 2007). To be conservative, we powered the study to detect the smallest effect observed in previous studies (Cohen’s f = .25; Goldenberg et al., 2005; Menzies & Dar-Nimrod, 2017; Menzies et al., 2021), which indicated that 128 participants were needed to reach 80% power ($\alpha$= .05).

Both studies employed a 2 (Fat-content; [High, Low]) x 2 (Prime; [Mortality, Dental Pain]) mixed-effects design to test the effect of death anxiety on portion size., which has two within-subjects levels (i.e., high fat, low fat). For the remaining dependent variables, we compared the two prime groups (i.e., mortality vs. dental pain), employing a standard between-subjects design. ED symptomatology was measured as a continuous moderator of the effect of the priming condition on ED behavior, separately in each study. The studies were approved by the University of Sydney Human Research Ethics Committee (Protocol: 2021/310) and pre-registered on Open Science Framework (https://osf.io/2m3vx/?view_only=8f2476b25dd6401da67ea3057d86df0e).

Study 1

Participants

We recruited undergraduate women at the University of Sydney, participating in exchange for course credit. In total, 132 students consented and 128 completed the study (97% completion). The mean age of participants was 19.69 years (SD = 3.21), and the mean EDE-Q score was 2.31 (SD = 1.40), which is moderately elevated compared to normative data for young women aged 18-22 years (Mond et al., 2006). A total of 20 participants reported a previous ED diagnosis (n = 14) and/or had sought treatment for an ED or related issues (e.g., poor body image; n = 15).

Measures and materials

Eating Disorder Examination – Questionnaire (EDE-Q) (Fairburn & Beglin, 1994) has 28 statements spanning four domains of ED symptomatology: restraint, eating concern, weight concern, and shape concern. Items are rated on a 7-point Likert scale from 0 to 6. The scale has been validated as a screening tool for EDs, and the overall scale has demonstrated good reliability (Aardoom et al., 2012). The EDE-Q had excellent internal consistency in both Studies 1 and 2 ($\alpha$s ≥ .92).

Salience Prime comprised two open-ended questions. Participants allocated to the MS priming condition answered: Please briefly describe the emotions that the thought of your own death arouses in you and Jot down as specifically as you can, what you think will happen to you as you physically die, and once you are physically dead (Arndt et al., 1998, p. 1218). The control condition employed the most well-validated stimuli; two similarly worded questions about dental pain (Burke et al., 2010; Long & Greenwood, 2013).

Big Five Aspects Scales (BFAS) neuroticism subscale (DeYoung et al., 2007) includes 20 items on a 5-point Likert scale (1 = strongly disagree; 5 = strongly agree). The neuroticism subscale has previously been shown to have good reliability (DeYoung et al., 2007). Internal consistency was questionable in Study 1 ($\alpha$= .69), but acceptable in Study 2 ($\alpha$= .72).

Positive and Negative Affect Schedule – Expanded Form (PANAS-X) (Watson & Clark, 1999) is a 60-item measure of current affect, with items rated on a 5-point Likert scale (1 = very slightly or not at all; 5 = extremely). The PANAS-X has been shown to have good reliability (Watson & Clark, 1999), and was included to confirm that there were no differences in affect following the primes. Internal consistency for both subscales was excellent in both studies ($\alpha$s ≥ .91).

Portion Size Estimation Task was a novel online portion size estimation task with a range of high- and low-fat snacks that we developed. We piloted the stimuli with 13 participants who rated the perceived fattiness of 17 snacks on a 7-point Likert scale. We selected three low-fat and three high-fat snacks for the task. The low-fat snacks (fat-content < 10%) had a subjective fat rating < 2, whereas the high-fat snacks (fat-content > 30%) had a subjective fat rating > 3.

Participants were presented with seven images of each of the six snacks that varied in portion size (see Figure 1). That is, all participants received images of three high-fat snacks and three low-fat snacks. Each image depicted a 20% increase in grams and the midpoint was the recommended serving size (i.e., 1 = smallest portion; 7 = largest portion). Participants were asked to imagine that they were eating the snack and indicate an appropriate portion size. The internal consistency of average portion size was questionable in Study 1 ($\alpha$= .68), but acceptable in Study 2 ($\alpha$= .70). The portion size task can be found at https://osf.io/gxtdk/?view_only=ea1f57580f5f4b1aa3b2216eb2e733db.

Figure 1. Example stimulus for novel portion size estimation task.

Female Body Scale (FBS) and Female Fit Body Scale (FFITBS) (Ralph-Nearman & Filik, 2020) are measures of body dissatisfaction that present nine female body figures in systematically increasing sizes. The figures are arranged based on adiposity (FBS) or muscularity (FFITBS). Participants indicated which figure best represents their current and ideal body figure. We calculated body dissatisfaction scores by subtracting participants’ current rating from their ideal rating.

Behavioral Intention was a single item designed to assess intention to eat. Participants were asked how willing they would be to participate in a follow-up study where they would sample different snacks, and responded using a 5-point Likert scale (1 = not at all; 5 = extremely).

Procedure

Recruitment occurred in August 2021. Participants responded to an advertisement for a study on the relationship between body image and eating behavior and read a cover story explaining that the research was being run in partnership with a snack foods company to conceal the aim of the study.

After providing demographic information, participants completed the EDE-Q and were randomized to one of two priming conditions, using the inbuilt Qualtrics function which assured anonymous randomization. To allow for sufficient delay between the prime and outcome measures, participants completed the neuroticism subscale of the BFAS and the PANAS-X before reading an article about marketing techniques and answering comprehension questions to bolster the cover story. Participants then completed the portion size estimation task, the FBS and FFITBS, and the behavioral intention item.

Finally, participants reported what they believed the study was about and whether they had been diagnosed with or sought treatment for an ED or related issues. Participants were then debriefed.

Results

Of the 128 participants who were randomized, seven failed an attention check, and one was under 18 and therefore excluded. The final sample comprised 120 participants with 58 allocated the MS prime and 62 the dental pain prime.

Prior to conducting the primary analyses, we used independent samples t-tests to assess for differences between the MS and control conditions on demographic variables and potential confounds. There was no significant difference between the priming conditions for age, ED symptomatology, levels of neuroticism, and positive and negative affect (ps>.159). As a result, we did not include covariates in the analyses of our dependent variables.

A one-way between-groups ANOVA revealed no significant difference in body dissatisfaction on the FBS, nor the FFITBS, for the MS prime compared with the control prime.

Unexpectedly, a one-way between-groups ANOVA found that behavioral intention was significantly greater for the MS prime (M = 3.17, SD = 1.37) than the control prime (M = 2.50, SD = 1.36), F(1,118) = 7.28, p = .008, η² = .06.

A 2 (fat-content) x 2 (prime) mixed-effects ANOVA found a main effect of fat-content, such that participants selected a larger mean portion size for low-fat (M = 3.14, SD = 1.04) than high-fat snacks (M = 2.59, SD = 0.97), F(1,118) = 37.79, p < .001, η_p² = .24. However, there was no significant main effect of prime, nor was there a significant interaction between the priming condition and fat-content.

We used Hayes (2017) PROCESS macro to test whether EDE-Q moderated the impact of the MS prime on body dissatisfaction, behavioral intention, and portion size. We conducted a hierarchical regression for each of the dependent variables in which we entered the priming variable and EDE-Q in the first step, and their interaction in the second step (see Tables 1 and 2). Unexpectedly, the interaction term failed to add to the variance in any of the regression models, indicating that ED symptom severity did not moderate the impact of the prime.

Table 1. Model summary predicting outcome measures from EDE-Q, prime, and their interaction.

	Female Body Scale	Female Fit Body Scale	Behavioral Intention	Portion Size
	Regression Model Summary
R^2	.35	.16	.06	.13
F	20.80	7.42	2.59	5.82
p	< .001	< .001	.056	.001
	Test of Highest-Order Interaction (PROCESS)
ΔR^2	.001	.021	.001	.005
ΔF	0.22	2.95	0.12	0.68
p	.642	.088	.733	.410

Note. N = 120. R² = Coefficient of Determination, interpretable as the proportion of variance in the outcome of interest which is explained by the predictors together. ΔR² = incremental improvement in the model offered by the interaction term.

Table 2. Unstandardized regression coefficients, and 95% confidence interval limits within models predicting outcome measures from EDE-Q, prime, and their interaction.

				95% CI
		b	SE(b)	LL	UL
Female Body Scale
	Constant	−1.12	0.12	−1.35	−0.89
	Prime	0.04	0.17	−0.29	0.37
	EDE-Q	−0.50***	0.09	−0.68	−0.32
	Interaction	0.06	0.12	−0.18	0.30
Female Fit Body Scale
	Constant	0.06	0.16	−0.26	0.37
	Prime	−0.10	0.23	−0.56	0.35
	EDE-Q	−0.20	0.13	−0.45	0.05
	Interaction	−0.29	0.17	−0.62	0.04
Behavioral Intention
	Constant	2.50	0.17	2.15	2.84
	Prime	0.68**	0.25	0.18	1.17
	EDE-Q	−0.03	0.14	−0.30	0.24
	Interaction	−0.06	0.18	−0.42	0.30
Portion Size
	Constant	2.89	0.11	2.68	3.10
	Prime	−0.05	0.15	−0.35	0.25
	EDE-Q	−0.17*	0.08	−0.33	−0.01
	Interaction	−0.09	0.11	−0.31	0.13

Note. N = 120. Prime; 0 = dental, 1 = death. EDE-Q = Eating Disorder Symptomatology. b = unstandardized regression weight. CI = Confidence Interval. LL and UL represent the lower and upper limits of the 95% confidence interval, respectively.

*p < .05; **p < .01; ***p < .001.

Contrary to the hypotheses, participants primed with death did not select smaller portion sizes, nor express greater dissatisfaction with their current thinness or muscularity compared to those primed with dental pain. The only difference between the two priming conditions was on the behavioral intention item, but the effect was in the opposite direction to the hypothesis. That is, participants in the MS priming condition were more willing to be followed up for a taste-test study than controls. Finally, the results did not support the hypothesized moderating effect of ED symptoms on the prime. Overall, there was no evidence that death anxiety influenced ED symptoms, as hypothesized, in this undergraduate sample.

Study 2

The design of Study 2 was identical to Study 1, except we recruited a clinically relevant sample of women who reported body image concerns or disordered eating behavior, and added the measures outlined below.

Participants

We recruited women using paid social media advertisements that were also distributed through ED organizations and support groups. Participants who completed the study could enter a draw to win a $75AUD gift voucher. In total, 300 women accessed and 276 consented to participating in the study. Of the 155 participants who reached the point of randomization, 129 completed the study (47% completion). We analyzed all available data for randomized participants. The mean EDE-Q score was at approximately the 90^th percentile (3.26; SD = 1.66) and above the clinical cutoff score (Department of Health, 2021; Mond et al., 2006), and 77 participants reported a previous ED diagnosis (n = 58) and/or had sought treatment for an ED or related issues (n = 74). The mean age of participants was 30.65 years (SD = 12.84).

Measures and materials

Study 2 used all the measures in Study 1, with the additions discussed below. We also measured participants’ subjective ratings of the tastiness and healthiness of the snacks. However, we excluded these scales from any inferential analyses because the Cronbach’s alphas were poor.

Hunger and Fullness was three items assessing the time of participants’ last meal and their current hunger and fullness on a 5-point Likert scale (1 = not at all; 5 = extremely). Consistent with prior research (e.g., Milos et al., 2013), we added these items to control for levels of satiety.

Existential Concerns Questionnaire (ECQ) death anxiety subscale (van Bruggen et al., 2017) includes 7 items rated on a 5-point Likert scale (1 = never; 5 = always). The reliability of this subscale has been shown to be good (Chawla et al., 2022; van Bruggen et al., 2017), and in the present sample the internal consistency was excellent ($\alpha$= .90).

Procedure

Recruitment occurred between July and September 2021. The procedure was identical to Study 1, except for the inclusion of the additional measures.

Results

A total of 155 participants were randomized. However, one participant was subsequently excluded for being under 18 years. Of the 154 remaining participants, 81 were allocated the MS prime and 73 the dental pain prime.

We conducted independent samples t-tests to determine whether demographic variables and potential confounds may have differed between the MS and control conditions. There was no significant difference between the priming conditions for age, ED symptomatology, levels of neuroticism, positive and negative affect, hunger and fullness, and levels of death anxiety (ps > .167). As a result, we did not include covariates in the primary analyses.

As predicted, a one-way between-groups ANOVA showed that FBS scores differed between the two priming conditions. Dissatisfaction with current thinness was greater for participants allocated the MS prime (M = −1.97, SD = 1.55) than controls (M = −1.41, SD = 1.39), F(1,126) = 4.62, p = .034, η² = .04. In contrast, scores on the FFITBS did not differ significantly between the conditions. A one-way between-groups ANOVA revealed no significant difference in behavioral intention for the MS prime compared with the control prime.

A 2 (Fat-content) x 2 (Prime) mixed-effects ANOVA revealed a significant main effect of fat-content, such that portion sizes were greater for low-fat (M = 2.72, SD = 1.05) than high-fat snacks (M = 2.18, SD = 1.04), F(1,128) = 34.92, p < .001, η_p² = .21. There was no main effect of prime. However, the interaction between prime and fat-content was significant (Figure 2), F(1,128) = 6.78, p = .010, η_p² = .05. Simple effects tests revealed that in the MS priming condition the difference in portion sizes for high versus low-fat snacks was significantly larger (M_D = 0.74, SE = 0.12), t(128) = 6.34, p < .001, than in the control condition (M_D = 0.29, SE = 0.13), t(128) = 2.23, p = .027. The other pair of simple effects showed that portion sizes for high-fat (p = .334) and low-fat foods (p = .138) did not differ between the prime groups.

Figure 2. Interaction effect between prime and fat-content on portion-size.

Note. Error bars represent standard errors. * p < .05; ** p < .01; *** p < .001.

We used Hayes (2017) PROCESS macro to examine moderation effects. We constructed separate hierarchical regressions for each of the dependent variables (i.e., body dissatisfaction, behavioral intention, portion size), entering the priming variable and EDE-Q in the first step followed by their interaction in the second step. None of the analyses demonstrated a significant interaction effect for any outcomes (see Tables 3 and 4), meaning ED symptom severity did not appear to moderate the impact of the MS prime, as hypothesized.

Table 3. Model summary predicting outcome measures from ede-q, prime, and their interation.

	Female Body Scale	Female Fit Body Scale	Behavioral Intention	Portion-Size
	n = 128	n = 128	n = 128	n = 132
	Regression Model Summary
R^2	.30	.08	.01	.11
F	17.68	3.38	0.41	5.41
p	< .001	.021	.748	.002
	Test of Highest-Order Interaction (PROCESS)
ΔR^2	.006	< .001	.004	.007
ΔF	0.98	0.02	0.44	0.99
p	.325	.892	.509	.321

Note. R² = Coefficient of Determination, interpretable as the proportion of variance in the outcome of interest which is explained by the predictors together. ΔR² = incremental improvement in the model offered by the interaction term.

Table 4. Unstandardized regression coefficients, and 95% confidence interval limits within models predicting outcome measures from EDE-Q, salience, and their interaction.

				95% CI
		b	SE(b)	LL	UL
Female Body Scale n = 128
	Constant	−1.46	0.17	−1.79	−1.13
	Prime	−0.51*	0.23	−0.95	−0.06
	EDE-Q	−0.38***	0.11	−0.59	−0.17
	Interaction	−0.14	0.14	−0.41	0.14
Female Fit Body Scale n = 128
	Constant	−0.09	0.22	−0.52	0.34
	Prime	−0.18	0.30	−0.77	0.41
	EDE-Q	−0.29*	0.14	−0.56	−0.02
	Interaction	0.02	0.18	−0.33	0.38
Behavioral Intention n = 128
	Constant	3.22	0.21	2.81	3.63
	Prime	−0.10	0.28	−0.66	0.45
	EDE-Q	−0.001	0.13	−0.26	0.26
	Interaction	−0.11	0.17	−0.45	0.23
Portion-Size n = 132
	Constant	2.36	0.11	2.15	2.57
	Prime	0.08	0.14	−0.20	0.37
	EDE-Q	−0.11	0.07	−0.25	0.02
	Interaction	−0.09	0.09	−0.26	0.09

Note. Prime; 0 = dental, 1 = death. EDE-Q = Eating Disorder Symptomatology. b = unstandardized regression weight. CI = Confidence Interval. LL and UL represent the lower and upper limits of the 95% confidence interval, respectively.

*p < .05; **p < .01; ***p < .001.

Overall discussion

In Studies 1 and 2, we examined whether a MS prime would exacerbate body image and ED related attitudes and behaviors in an undergraduate and clinically relevant sample, respectively. In the undergraduate sample, results were inconsistent with a causal role of death anxiety in body image issues and disordered eating. However, in the clinically relevant sample, MS priming led to greater dissatisfaction with current thinness, but not muscularity. There was also a significant interaction between priming condition and the fat-content of the snacks on the portion size estimation task. However, there was no effect of MS priming on intention to eat the snacks in a follow-up study. In both the undergraduate and clinically relevant sample, unexpectedly, ED symptom severity did not moderate the relationship between MS priming and body image and ED related outcomes.

Taken together, the above findings suggest that death anxiety may drive women to adhere to cultural standards promoting thinness in an effort to bolster self-esteem, but only amongst those who already report problematic body image or eating behavior. In the clinically relevant group, MS priming resulted in greater dissatisfaction with current thinness and restriction of portion sizes of high-fat compared to low-fat snacks. That is, reminders of death led participants to perceive themselves as further from their ideal thinness and motivated them to minimize portions of high-fat relative to low-fat foods. This finding is in line with research showing that women with EDs prefer low-fat food and specifically restrict calories from fat (Gianini et al., 2019; Steinglass et al., 2015). Overall, these results provide some support for the potential role of death fears in ED attitudes and behavior, particularly those in which the pursuit of thinness is a central concern.

Contrary to our hypothesis, the MS prime did not impact dissatisfaction with current muscularity. Muscularity is generally an attribute that is more culturally important for men (Griffiths et al., 2013), which may be why the women in the present study did not display muscle-related body dissatisfaction when reminded of death.

We also failed to find that MS priming reduced behavioral intention, as hypothesized, in either study. In fact, in the sample of undergraduate students, we found the opposite effect. It is possible that by situating the “taste-test” within the context of a “follow-up study” the salient behavior shifted from eating to participating in research. If participants interpreted this item to be about future research rather than eating, then there may have been no need for the avoidance behavior we anticipated in the clinically relevant sample. Furthermore, in line with TMT arguments, the undergraduate psychology students might have derived self-esteem from volunteering to participate in research because it is important to their worldview. This rationale could explain why death anxiety led to increased behavioral intention in Study 1 but failed to produce any effect in Study 2.

Neither study found evidence that ED symptom severity moderated the impact of the prime on any outcomes. It is possible that by sampling two populations with different levels of ED symptomatology, there was insufficient variability within each sample, reducing the power to detect a significant moderation effect. Nonetheless, the fact that a MS priming effect was found in the clinically relevant sample, but not in the undergraduate sample suggests that death anxiety has a more pronounced impact on body image and disordered eating among women with these concerns. This finding is consistent with the notion that death fears selectively exacerbate disorder-consistent coping strategies (e.g., Menzies et al., 2021a; Menzies & Dar-Nimrod, 2017).

Although the results of our studies are mixed, the findings suggest that a brief death anxiety manipulation increased the pursuit of thinness among women with problematic body image or eating behavior. These results provide some support for the claim that death anxiety is a transdiagnostic construct relevant to symptoms of EDs (Goldenberg et al., 2005; Iverach et al., 2014; Le Marne & Harris, 2016). Finding potential psychological mechanisms that contribute to the development of EDs is important considering these disorders are amongst the most severe mental illnesses (Udo & Grilo, 2019; van den Berg et al., 2019). Further, current psychological treatments have not been shown to provide significant, long-lasting benefit for anorexia nervosa (van den Berg et al., 2019) and the rate of relapse for anorexia nervosa remains high (Khalsa et al., 2017).

The present findings suggest that death anxiety could be a therapeutic target for women with body image issues and EDs characterized by the pursuit of thinness. This theoretical idea is worthwhile exploring further in future research given the modest outcomes of available treatments for anorexia nervosa (Solmi et al., 2021). As cognitive-behavioral approaches have been shown to reduce death anxiety (see Menzies et al., 2018), interventions targeting fears of death may improve the outcomes of ED treatments. Future research could investigate whether such interventions provide additive benefits to standard treatments for EDs.

Despite careful attention to methodology, some limitations should be noted. First, we developed the portion size estimation task as a novel measure of disordered eating that could be administered remotely. Although it was designed as a proxy measure of eating, it contrasts to previous laboratory research, involving actual eating, which might be more sensitive to MS priming effects (Goldenberg et al., 2005). Second, it is unclear how representative the clinically relevant sample was of people seeking treatment for EDs. Many participants in Study 2 reported an ED diagnosis and/or having sought treatment for ED related issues, but this was not independently verified. Further, the completion rate in Study 2 was quite low (i.e., 47%). Nevertheless, the clinical relevance of the sample is supported by the high ED symptom scores observed (i.e., 90th percentile on EDE-Q; Mond et al., 2006). Further, the mean score on the ECQ death anxiety subscale exceeded that observed in a treatment-seeking OCD sample (Chawla et al., 2022). Therefore, we can be reasonably confident that it was a clinically relevant sample.

Negative self-evaluation related to body weight and shape and the idealization of thinness are common to anorexia nervosa, bulimia nervosa, and binge-eating disorder (5th ed.; DSM-5; American Psychiatric Association, 2013) and so we did not distinguish between these presentations in the clinically relevant sample. As death anxiety is claimed to be transdiagnostic, we would argue that it is a strength of the study that it included a heterogenous sample. Nonetheless, future research could investigate whether death anxiety impacts individuals diagnosed with restrictive EDs differently from those characterized by bingeing and purging.

Despite the extensive literature demonstrating that death anxiety affects a broad repertoire of behavior, few studies have investigated its role in mental illness. Existing research has largely focused on disorders where there is an overt preoccupation with illness or death (Menzies et al., 2021a; Menzies & Dar-Nimrod, 2017). Hence, our work is among the first to provide experimental support for the theory that death anxiety drives symptoms of disordered eating, at least in a clinically relevant sample of women.

In these wellpowered, pre-registered studies, we demonstrated that MS priming exacerbates specific ED attitudes and behavior consistent with the “thin ideal,” but only among women with body image concerns or disordered eating behavior. The present studies contribute to mounting evidence in support of Iverach et al. (2014) claim that death anxiety is a transdiagnostic construct. The results suggest that EDs could be another diagnostic group in which symptoms are, at least partially, driven by death anxiety. This may mean that considering death anxiety in the conceptualization of EDs and targeting it in treatment could result in symptomatic improvement. However, this currently remains speculative and further research is required for this to be confirmed.

Disclosure statement

The authors report there are no competing interests to declare.

Data availability statement

Data that support the findings of the studies will be made available by the corresponding author upon reasonable request.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.