Shame as a mediator of the association of childhood emotional abuse with aversive cognitive perseveration in adults

Abstract

Childhood emotional abuse (CEA) has been linked to response‐focused emotion regulation in adulthood. However, the underlying mechanisms remain unexplored. This pre‐registered study examined whether shame mediates the association between CEA history and aversive cognitive perseveration (ACP), including brooding rumination, experiential avoidance and emotional non‐acceptance, in adulthood. Ninety‐three undergraduates (66.6% female; age: 21.62 [7.63]) completed self‐report measures of maltreatment history, trait shame and trait ACP and underwent a piloted shame induction task using a within‐subject design. The assessment involved baseline and post‐induction reactive‐state shame and negative affect, subsequent use of reactive‐state ACP and engagement strategies, followed by childhood maltreatment and trait measures. Two sets of state and trait mediation models were tested using Bayesian estimation, adjusting for reactive‐state negative affect, reactive‐state engagement strategies and the sum of physical abuse, emotional neglect and physical neglect histories. The models primarily supported the association of shame with ACP across state and trait models. CEA history was associated with reactive‐state and trait ACP through the indirect effect of trait (but not reactive‐state) shame. A post‐hoc moderation analysis suggested reduced post‐induction shame and ACP use for moderate‐to‐severe CEA history, which was tentatively attributed to the involvement of aberrant stress reactivity associated with higher CEA history.

BACKGROUND

Childhood emotional abuse (CEA), which entails caregivers' non‐physical maltreating behaviours, such as spurning, terrorizing and threats of harm to the child, has recently shown a higher prevalence rate compared to other maltreatment types, especially physical and sexual abuse (Baker, 2009; Bernstein et al., 2003; Mathews & Dube, 2025). Importantly, recent meta‐analytic studies on quasi‐experimental, prospective and retrospective associations have underscored the significance of CEA in a wide range of psychopathologies and the associated risk factors (Baldwin et al., 2023, 2024; Li et al., 2020; Palmier‐Claus et al., 2024; Xiao et al., 2023). Specifically, CEA history is linked to elevated emotional reactivity and emotion regulation difficulties in adulthood (Amini‐Tehrani et al., 2025; Gratz et al., 2007; Shapero et al., 2014; Wang et al., 2013). However, research has yet to pinpoint a transdiagnostic emotion regulation mechanism to explain the diverse consequences in adult survivors of CEA.

Emotion regulation is a multifaceted construct through which individuals use behavioural or cognitive processes, particularly emotion regulation strategies, to attain their affective and situational goals (Aldao, 2013). Importantly, greater response‐focused strategy use, that is, modification of the emotional trajectory after the emotion is fully elicited (Gross, 2015), is proposed to underlie the association of CEA history with adult psychopathology (Weissman et al., 2019). Among diverse response‐focused strategies, the current study investigates strategies characterized by self‐referential thinking (Mennin & Fresco, 2013) because CEA history has been consistently linked to elevated self‐referential processing (Krause‐Utz et al., 2014; Wang et al., 2013).

Specifically, a recent meta‐analytic study investigating eleven emotion regulation constructs (Amini‐Tehrani et al., 2025) highlighted associations of CEA history with trait use of negative self‐referential thought patterns embodied in brooding rumination (i.e., self‐reproaching thoughts and focus on causes and consequences of negative mood, feelings and behaviours; r = 0.37) (Treynor et al., 2003), experiential avoidance (i.e., effortful attempts to avoid unwanted internal experiences, such as thoughts, emotions and bodily sensations; r = 0.30) (Hayes et al., 1996) and emotional non‐acceptance, particularly in the form of developing secondary emotions to primary feelings (i.e., non‐accepting meta‐emotions, r = 0.26) (Gratz & Roemer, 2004; Mitmansgruber et al., 2009). Lab‐based studies have also demonstrated that the use of non‐accepting meta‐emotions in response to distressing stimuli could explain avoidant tendencies associated with CEA history (Gratz et al., 2007). Importantly, although CEA history showed a generally small meta‐analytic association with generally adaptive reappraisal tendencies (Amini‐Tehrani et al., 2025) (i.e., changing the meaning of stimuli to reduce negative or positive emotions, r = −0.08) (Gross, 2015), research has suggested that CEA history may nonetheless be associated with reinterpretation difficulties due to alterations in semantic networks involved in self‐referential processing (Lee et al., 2021). However, the specific emotional context that encourages the use of those self‐referential emotion regulation strategies has remained unexplored.

Research and theory suggest that in a given emotional episode, individuals may implement multiple strategies simultaneously (Sheppes et al., 2015). Consistent with this, aversive cognitive perseveration (ACP) emerged as a broad trait strategy factor through meta‐analytic intercorrelations between brooding rumination, experiential avoidance and emotional non‐acceptance, which together represent “over‐engagement in negative cognitions while rejecting and wanting to avoid negative feelings and emotions” (Naragon‐Gainey et al., 2017, pp. 409–410). Additionally, studies of momentary strategy use in daily life have generally supported the ecological validity of the ACP strategy factor, individuals tend to use ACP components closely in daily emotion regulation episodes to deal with distress (McMahon & Naragon‐Gainey, 2019). Empirically, recent studies have also supported that ACP forms a coherent cognitive process that operates as an over‐engagement emotion regulation tendency. For example, ACP longitudinally mediated the association between intolerance of uncertainty and anxiety (Sahib et al., 2024), and it was associated with neuroticism (Daniel et al., 2024) and conflict resolution difficulties underlying diminished family functioning (Low et al., 2019).

ACP operates on the response‐focused emotion regulation stage, providing an avoidant function. That is, ACP thought patterns sustain critical attention on emotionally charged self‐states, brooding over the loss of desired self‐states (‘why can't I do better?’), suppressing the emerging negative self‐states (e.g., ‘I need to make my feelings go away’) and affectively rejecting these self‐states (e.g., ‘I'm angry that I'm feeling distressed’). The avoidant function of ACP thought patterns could be attributed to intense self‐processing aimed to rectify perceived discrepancy with self‐ideal standards (e.g., ‘why I couldn't respond in certain way’) (Watkins, 2004). In doing so, the individual acutely disavows the emerging negative self‐states (e.g., ‘what has manifested is not me’) (Eterović, 2020). Such thought patterns may serve a function of momentary relief from the affective loads of negative situations (e.g., a failure moment) (Hayes et al., 1996; Nolen‐Hoeksema et al., 2008; Watkins & Roberts, 2020).

To account for the role of CEA history in individual differences in ACP use in adulthood (Diaz & Eisenberg, 2015), we tested a mechanism in which CEA history operates through shame‐based self‐representations, which subsequently elicit ACP use. Specifically, shame has been conceptualized as a complex emotion involving cognitive appraisals of one's devalued self (Andrews et al., 2002; Landers et al., 2024; Lewis, 1987). Notably, shame as a chronic trait has shown a large effect size with diminished trait self‐esteem (Budiarto & Helmi, 2021), dysfunctional guilt‐proneness in symptomatic samples (Kim et al., 2011; Possemato et al., 2021; Shi et al., 2021) and a diverse range of symptomatology in adulthood (Allan et al., 1994; Azevedo et al., 2022; Laving et al., 2023; Pallanti & Quercioli, 2000; Sheehy et al., 2019; Shi et al., 2021; Tangney et al., 1992). Importantly, CEA history reflects early instances of humiliation and shaming by attachment figures (Farr et al., 2021; Gilbert et al., 1996), which directly affect the child's self‐worth (Maguire et al., 2015; Riggs, 2010). The association of CEA history with shame tendencies in adulthood has been embodied in various ways and research contexts. Of note, meta‐analytic studies have shown moderate‐sized associations between CEA history and trait shame (Zhang et al., 2025) and defectiveness shame schema (Pilkington et al., 2021), which could generally exceed those observed for other maltreatment types (e.g., emotional neglect). In the lab, CEA history has been associated with higher shame‐based self‐views, especially from one's own perspective than other's perspective (Wever et al., 2023). In the same vein, CEA history was associated with negative self‐referential processing in response to lab‐based self‐esteem threatening situations (van Harmelen et al., 2011), and adults with CEA history may tend to avoid processing induced rejection cues from others (Hoepfel et al., 2025). In light of the increased prevalence rates, these trait and state associations suggest that CEA history may identify a large number of adults with shame‐related difficulties.

Moreover, empirical research has supported the trait associations between shame and greater use of ACP components, including trait rumination (Cenkner et al., 2023; Peters et al., 2014), trait experiential avoidance (Carvalho et al., 2015; Leonard et al., 2020) and trait non‐accepting meta‐emotions (Mirabile et al., 2023). The state association between shame and ACP components, especially rumination, in reaction to self‐esteem‐threatening stimuli has also been frequently supported in the lab (Cândea & Szentágotai‐Tătar, 2017; Milia et al., 2021; Vrshek‐Schallhorn et al., 2019). Current literature also supports associations across CEA‐associated features, trait rumination, trait shame and their interrelations (e.g., borderline features, Peters et al., 2014). However, the role of shame in linking CEA to ACP use has not yet been tested. Additionally, only limited research has explored CEA history in terms of spontaneous emotional reactivity and regulation in the lab (Gratz et al., 2007). Overall, it is plausible that shame may specifically function as an emotion context (Aldao, 2013) in which response‐focused tendencies associated with CEA history could operate through ACP use in response to shame.

To operationalize shame with respect to tendencies associated with CEA history, we drew on the self‐attribution theory of shame (Tangney et al., 1992; Tracy & Robins, 2004) and the constructionist theory of emotions (Barrett, 2016), which respectively explain the mechanism of shame reactions and that of subjective experience and expression of shame. Specifically, shame reactions may generally occur when individuals ascribe negative events to their entire self or a stable character (Tangney et al., 1992). In fact, the self‐attribution mechanism of shame fundamentally operates on negative self‐representations (Tracy & Robins, 2006). Additionally, according to the constructionist theory of emotions, state emotional experience generally entails cognitive self‐categorization via diverse concepts that are representative of the varying shades of the current affective experience (Barrett, 2016; Barrett et al., 2007). In tandem, these theories suggest that, instead of a global ‘shame concept,’ shame experience may consist of a wider range of affective experiences that can tap into multiple aspects of self‐representations that unfold along the shame experience.

For example, self‐representations can be sourced from general negative beliefs about oneself and how negatively oneself is represented to others, in addition to a negative self‐evaluative valence (Thagard & Wood, 2015). In terms of subjective shame experience, these self‐representations generally deal with a sense of awkwardness, inadequacy, ridicule and expression of vulnerability to exposure (Harder et al., 1993; Lewis, 1987). Importantly, while trait shame entails frequent reactivity to self‐esteem threatening cues across situations, state shame reactions are generally graded in intensity in a manner that manifests in increasingly harsher self‐descriptors and expressions in a given situation (see Garofalo et al., 2025, for a review of measurement diversity). Accordingly, we operationalized state shame in the lab based on affective self‐descriptors on a continuum of increasing harshness in tone (e.g., ‘feeling childish’ to ‘feeling disgusting to others’), in addition to increasing harshness in expressions (e.g., ‘feeling of blushing’ to ‘feeling helpless’) (Harder et al., 1993; Harder & Zalma, 1990; Turner, 2014). We also operationalized trait shame based on multidimensional negative self‐representations across behaviours, character and appearance in the eyes of oneself and others (Andrews et al., 2002; Garofalo et al., 2025). Together, these operationalizations are also conceptually consistent with the severity of harm to self‐worth as a key childhood impact of CEA (Maguire et al., 2015; Riggs, 2010).

Accordingly, we hypothesized that CEA history would be associated with higher rates of shame‐based self‐representations in response to a shame induction, which in turn would predict both reactive‐state and trait use of ACP. To investigate this hypothesis, we tested two pre‐registered mediation models, including the reactive‐state shame model predicting reactive‐state ACP use (Figure 1A) and trait shame model predicting reactive‐state (Figure 1B) and trait ACP use (Figure 1C). Furthermore, to isolate the unique effect of CEA history and shame, we conducted a set of pre‐registered secondary analyses. Specifically, to isolate the associations for CEA history, we further adjusted the primary models by the composite score of physical abuse, physical neglect and emotional neglect (i.e., adjusted score of general maltreatment, see Ethical Considerations and Limitations for details regarding the omission of sexual abuse history). Additionally, the models that included reactive‐state ACP were also adjusted for state use of engagement emotion regulation strategies (e.g., positive reappraisal) to account for resilience factors that could reduce emotional reactivity (Szentagotai‐Tatar & Miu, 2016; Zhang et al., 2020). Importantly, given that self‐esteem threatening situations can elicit broader negative affect, the models that included reactive‐state ACP were also adjusted for state negative affect to isolate the unique effect of shame.

FIGURE 1.

Pre‐registered conceptual models. State Emotion regulation‐T2 includes the use of aversive cognitive perseveration (rumination, experiential avoidance and non‐acceptance of emotion). Panels (A) primary model, (B) secondary mediation model of trait shame and (C) secondary mediation model of trait shame and trait ACP. The models are further adjusted by (a) state use of constructive strategies, (b) use negative affect, (c) the sum of other maltreatment types (physical abuse, physical neglect and emotional neglect).

Moreover, it was plausible that these observed effects were different for the facets of ACP, given that ruminative thought patterns operate in a longer timeframe while experiential avoidance and non‐accepting meta‐emotions operate acutely (e.g., Grafton & MacLeod, 2013). Thus, we conducted a non‐registered sensitivity analysis that retested the reactive‐state models for the use of ACP components in the lab. We argued that a similar pattern of results would generally consolidate reactive‐state ACP use as coherent thought patterns in response to self‐referent stimuli.

METHODS AND MATERIALS

Design and sample

The current study aims, methods and analysis plan were pre‐registered at https://osf.io/nx6uy/?view_only=668d4e7b11d148c6b7f47b11d881b61b (anonymized link). This study was comprised of (1) a piloting phase utilizing a randomized between‐subject design to validate the novel shame induction paradigm and (2) a primary experiment utilizing a non‐randomized within‐subject design. The experiment schedule is detailed in the Supplementary Material Figure Supp‐I, where the time elapsed between assessment points and the administration of each scale is annotated.

For the pilot and primary study combined, the study sample consisted of 131 participants enrolled in an undergraduate psychology course. Eligibility criteria were (1) 18–65 years old and (2) score < 21 on the Depression subscale of Depression, Anxiety, Stress Scale (DASS‐21) representing normal to moderate depression levels (Lovibond & Lovibond, 1995). Exclusion of individuals with higher depressive symptoms was pre‐registered due to ethical concerns regarding the shame induction (i.e., reducing risk of harm if severely depressed individuals were included – see the Limitations section for further discussion).

From the enrolled participants, 23 participants completed the piloting phase that tested the validity of the shame induction procedure (see Supplementary Material‐Piloting). Furthermore, among the 108 participants completing the primary experiment, data were missing for two participants due to a computer error. From the rest (n = 106), eight participants were removed based on an expression of disbelief in the debriefing section, and one participant indicated inattentiveness in all the attention check items (see Supplementary Material‐Response Validity Items). Manipulation checks described below in the remaining data (n = 97) showed that two participants failed in the core items and one more failed in the total items (who also expressed a minor hypothesis guessing in the debriefing – see Supplementary Material‐Manipulation Check). We also examined multivariate outliers using Mahalanobis distance, which resulted in the removal of one more participant (p < .001). Therefore, 93 participants remained in the final analysis.

Supplementary Material Figure Supp‐II reports the flow of participations and Table Supp‐I reports the sample characteristics. Briefly, the final sample for the primary study (N = 93) had a mean age of 21.62 (standard deviation [SD] = 7.63, range:18–63; 66.6% female). The analysed sample showed an adequate range of depressive symptoms reported in‐session, with a mean DASS‐21 depression of 11.39 (SD = 3.91), ranging across normal (n = 39, 41.9%), mild (n = 31, 33.3%) and moderate levels (n = 22, 23.7%), in addition to one participant who reported slightly severe depressive symptoms in the lab (1.0%, despite being eligible based upon their lower depression level at screening). In the same vein, CEA history was distributed across all possible ranges, including none to minimal (n = 45, 48.4%), low (n = 26, 28.0%), moderate (n = 10, 10.8%) and severe levels (n = 12, 12.9%) (CTQ‐SF; Bernstein et al., 2003). No significant differences were observed between the analysed and removed participants in the demographic and clinical variables.

Procedure

To reduce demand characteristics and maintain the study objectives, we adopted a concealed experiment narrative in which the study was advertised as examining ‘individual experience in group activity.’ The procedure described in Supplementary Material Figure Supp‐I and elaborated in Supplementary Material‐Assessment Schedule and Instruction was conducted on the Qualtrics platform using a computer during an in‐person session in a testing room. The briefing provided a concealed information sheet, a concealed consent form and an oral presentation of standardized instructions for the computerized group activity. The session was concluded with a sensitive debriefing described in detail in the Supplementary Material‐Debriefing.

Shame induction paradigm

The Supplementary Material provides the details of the shame induction paradigm, manipulation check, the filler task and the briefing procedure. Briefly, we devised a group activity embedded in a modified version of the Cyberball paradigm – a ball tossing activity on a computer (Williams & Jarvis, 2006). Cyberball has been shown to effectively induce shame in the original social exclusion paradigm (Catton et al., 2025). In our modified version, individual participants joined three allegedly real players who were actually computerized avatars to conduct a collaborative group activity that was designed to terminate with a brief social exclusion (i.e., no ball passed to the real participant), followed by a direct message of removal due to the actual participant's lowest performance in the group. This modified set‐up could render several aspects of shame elicitation and regulation in terms of self‐relevance, performance standard and spontaneous emotion regulation. Importantly, the shame reactivity was assessed at the induction point (T1 – roughly 7.5 min since the activity commenced), whereas ACP was assessed for the period elapsed between T1 and T2 following a filler task (roughly 5 min since the post‐induction assessment commenced, see Supplementary Material Figure Supp‐I). Specifically, we bridged the post‐shame induction (T1) and emotion regulation (T2) assessment points using a filler task, given that individuals generally tend to avoid self‐focused attention to minimize distress (Eterović, 2020). The filler task required selecting the image of a favourite object to allow time for self‐referential reflection on benign self‐aspects, maintaining the integrity of the underlying self‐focused attention. The induction section concluded with a battery of questionnaires assessing childhood maltreatment history, trait emotion regulation, the DASS‐21 Depression and Anxiety subscales and demographics.

To examine the internal validity of the shame induction, the pilot control group completed the same Cyberball task that excluded the active effect of the warning and the removal messages. However, the control Cyberball task retained the brief social exclusion end stage to induce a minor discomfort to avoid potential gamification that could contradictorily induce a social inclusion effect (e.g., Catton et al., 2025). The piloting experiment supported the expected shame response in the experiment group (n = 10) as compared to the control group (n = 11) (F = 9.65, p < .01, eta‐squared = 0.35, Observed power = 0.84; see Supplementary Material‐Piloting). Additionally, shame increased in 90% of the experiment group (overall change score range: 0 to +16), compared to only 36% of the control group (overall range: −3 to +5; χ ²[1, N = 21] = 6.88, p = .009, see Table Supp‐II).

Measures

Details of psychometric properties and adapted items of trait assessments are provided in the Supplementary Material‐Primary Measure, ‐Secondary Measures and ‐Filler Questions, where the ‐Assessment Schedule and Instruction section elaborates on the specifics of assessment requirement and schedule. Notably, the state scales/items had to be minimally altered in wording to align with the study context and the flow of assessment, especially in terms of timing (e.g., right now) and the descriptive response options (e.g., Likert anchors). The Cronbach's alphas for almost all the trait and state scales and subscales range between 0.70 and 0.94 in the current sample, representing adequate to strong internal consistency with the exception of state shame in the pre–manipulation‐T0, which showed low internal consistency (α = 0.62; see Table 1).

TABLE 1.

Correlation matrix between study variables (N = 93).

Variables	Cronbach's α	M	SD	Sk	Kr	1	2	3	4	5	6	7	8	9
1. CEA	0.83	10.03	4.45	1.19	1.05	1
2. Baseline‐state shame‐T0	0.62	12.78	2.64	1.61	3.06	.12	1
3. Reactive‐state shame‐T1	0.85	19.53	6.45	0.92	0.38	−.14	.33**	1
4. NA‐T1	0.74	10.01	3.31	1.01	0.60	.02	.16	.64***	1
5. Reactive‐state ACP‐T2	0.91	26.91	9.78	1.02	0.59	.07	.30**	.53***	.47***	1
6. Reactive‐state engagement strategy use‐T2	0.78	16.55	5.49	0.42	−0.33	−.11	.23*	.11	.10	.31**	1
7. Trait shame	0.94	57.72	15.04	0.09	−0.53	.36***	.33**	.31**	.26*	.45***	.01	1
8. Trait ACP	0.88	42.06	16.32	0.08	−0.61	.29**	.23*	.40***	.37**	.63***	.04	.69***	1
9. General maltreatment	0.86	22.62	7.47	0.86	0.03	.80***	.22*	−.06	.05	.15	−.04	.33**	.33**	1

Note: General maltreatment is the sum of physical abuse, emotional neglect and physical neglect.

Abbreviations: ACP, Aversive cognitive perseveration, encompassing brooding rumination, experiential avoidance and non‐accepting rejective emotions (rescaled to 0–100 scale format); CEA, childhood emotional abuse; Kr, Kurtosis; M, mean; NA, negative affect; SD, standard deviation; Sk, skewness.

^* p < .01.

^** p < .01.

^*** p < .001.

Briefly, we assessed CEA history via the Emotional Abuse subscale from the Childhood Trauma Questionnaire‐Short Form (CTQ‐SF), and we defined the adjusted score of general maltreatment as a composite score of Physical Abuse, Physical Neglect and Emotional Neglect subscales (Bernstein et al., 2003). We also assessed trait shame via the Experience of Shame Scale (ESS), trait rumination via the Brooding subscale from the Ruminative Response Scale (RRS)–10 (Treynor et al., 2003), trait experiential avoidance via the Brief Experiential Avoidance Questionnaire (BEAQ; Gámez et al., 2014) and non‐accepting meta‐emotions via Non‐Acceptance subscale of the Difficulties in Emotion Regulation Scale‐36 (DERS‐36) (Gratz & Roemer, 2004). Due to different scaling formats, we rescaled the three trait ACP components using percent of maximum possible scores technique into a 0–100 scale (Cohen et al., 1999).

Additionally, we assessed baseline‐ and reactive‐state shame via 10 affective words derived from the shame subscale of the Personal Feelings Questionnaire‐2 (PFQ‐2) (e.g., feeling stupid, humiliated, embarrassed and helpless) (Harder & Zalma, 1990). We also assessed reactive‐state negative affect via six negative affect items (Distressed, Afraid, Upset, Nervous, Hostile and Scared) from the Positive and Negative Affect Schedule – Short Form (PANAS‐SF; Mackinnon et al., 1999). The reactive‐state ACP components were measured via 1. Seven adapted items from the Brief State Rumination Inventory (BSRI; Marchetti et al., 2018) for reactive‐state rumination; 2. Three adapted items from Kashdan et al. (2013) and Levin et al. (2018) in addition to one revised item from the Brief Experiential Avoidance Questionnaire (BEAQ; Gámez et al., 2014) for reactive‐state experiential avoidance and 3. Four items from the State DERS (S‐DERS) for non‐accepting meta‐emotions. We particularly omitted the meta‐emotions of shame, embarrassment and guilt from S‐DERS to maintain discriminant validity from shame (Lavender et al., 2017). Finally, to assess reactive‐state engagement emotion regulation strategy use, we selected seven items suitable to the current emotional context (e.g., positive reappraisal) from the Cognitive Emotion Regulation Questionnaire (CERQ) – Short Form (Garnefski & Kraaij, 2006) (see Supplementary Material‐Table Supp‐III for adapted items).

Ethical considerations

The study received ethics approval from the Human Research Ethics Committee of the host university (Ref. 2023/ET000834), adhering to the tenets of the Australian National Statement on Ethical Conduct in Human Research (NHMRC, 2023). Confidentiality was maintained and the data were collected anonymously. Participants' welfare was prioritized, and written informed consent was sought before the commencement of the experiment (i.e., concealed informed consent) and after the disclosure of study aims (i.e., disclosed informed consent). No participant declined the disclosed informed consent. Specifically, we excluded the Sexual Abuse subscale of CTQ‐SF from the survey block due to the emotional impact frequently expressed in the debriefing stage for initial participants. Limitations are discussed in the Discussion section.

Data analysis

Registered analysis plan

We used SPSS for preliminary analysis (e.g., data cleaning, pilot study) and Mplus software version 8.11 (Muthén & Muthén, 2017) to test the pre‐registered primary and secondary models. The structural validity of state ACP‐T2 as a single factor was examined by confirmatory factor analysis (CFA) using the Maximum Likelihood with Robust Standard Errors (MLR) estimator, which accounts for non‐normality and missing data with Full Information Maximum Likelihood (FIML, n = 1). The Comparative Fit Index (CFI) (>0.90 for acceptable and >0.95 for excellent fit), the Root Mean Square Error of Approximation (RMSEA) and Standardized Root Mean Square Residual (SRMR) (≤0.08 for acceptable fit) indicated goodness of model fit (Browne & Cudeck, 1992; Hu & Bentler, 1999).

Path analysis was used with the Bayes estimator, which appropriately handles non‐normal data and missing data, assuming missing at random. We set the analyses for at least 10,000 iterations to ensure converging posterior scale reduction (PSR around 1.01) and valid bootstrapped credibility intervals (CrI) for indirect effects (not available with MLR estimator). We employed the default non‐informative priors that handle Bayesian posterior distributions using the observed data (Muthén & Asparouhov, 2012). Mediation model fit was evaluated using Bayesian CFI and RMSEA indices as per the cut‐off points described above. Statistical significance was determined by an alpha of 0.05 and a 95% CrI that excludes zero.

Non‐registered analysis plan

Following the pre‐registered analyses, we decided to conduct several follow‐up mediation and moderation analyses to account for observations. Specifically, we conducted a post‐hoc moderation analysis that inspected the tentative moderation effect of CEA history in the association of reactive‐state shame‐T1 and ACP‐T2 in the lab (see the results for follow‐up justification). In the moderation analysis, we employed the Bayes estimator with the GIBBS algorithm to render calculations and 70,000 iterations for PSR around 1.01, which yields a valid 95% CrI. We probed a possible significant interaction using the Johnson–Neyman plot provided by the CAHOST Excel workbook (Carden et al., 2017). In the moderation analysis, CEA and reactive‐state Shame‐T1 scores were grand mean centred.

Furthermore, we conducted a post‐hoc mediation analysis in which the primary state mediation models were further adjusted by trait shame (see the results for follow‐up justification). Finally, the same registered and non‐registered analysis plan was set for the ACP‐focused sensitivity analyses, in which the adjusted versions of the primary, secondary and post‐hoc mediation and moderation models were retested for separate ACP components substituted independently for the total score of reactive‐state ACP‐T2. Model fit for these post‐hoc models was appraised as per the registered analysis plan, and the respective results are provided in the Supplementary Material.

Power analysis

The power analysis suggested that a sample size of 105 suggested a power of 0.80 with a 95% confidence interval, assuming a correlation of .35 between variables (Amini‐Tehrani et al., 2025; Cenkner et al., 2023; Kouri et al., 2023; Leonard et al., 2020; Zhang et al., 2025). Data collection resulting in 106 participants for the primary analysis was terminated at the planned time (November 2024), and the final sample included 93 participants. Using the same effect size estimates described above, a sample of 93 people would produce a power of .67. We discuss the associated sample size limitations in the discussion section.

RESULTS

Preliminary analyses

Paired‐sample t‐tests indicated a significant increase in state negative affect‐T1 (Δmean = 1.62, SD = 3.06, p < .001, Cohen's d = 0.56) and shame reactivity‐T1 (Δmean = 6.74, SD = 6.12, p < .001, Cohen's d = 1.28), compared to the baseline state‐T0. These results suggest that the shame induction was successful.

Factor structure of aversive cognitive perseveration

We tested two CFA models of reactive‐state ACP to determine the best fitting representation, prior to the main analyses. These models include Three Factors on which three covaried item‐level factors for brooding rumination (seven indicators), non‐accepting meta‐emotions (four indicators) and experiential avoidance (four indicators) were loaded, and One Factor on which all the 15 ACP indicators were loaded simultaneously. As provided in Table 2, the Three Factor model indicated a borderline to acceptable fit, with valid factor loadings for Rumination: 0.436–0.809; Experiential avoidance: 0.641–0.723; Non‐accepting meta‐emotions: 0.342–0.731, p < .001. However, the factor correlations exceeded a value of 0.93, suggesting highly overlapping state constructs (α range: 0.70–0.84 & Ω range: 0.74–0.83).

TABLE 2.

Model fit indices.

	Reference Table/Figure	χ 2 [df]	p‐value (PPP)	CFI	TLI	RSMEA	p *	SRMR
CFA on the Reactive‐State ACP‐T2 using MLR estimator
Three‐factor	Table Supp‐III	134.9 [87]	<.001	0.894	0.872	0.077	.049	0.068
One‐factor		136.5 [90]	<.01	0.897	0.880	0.078	.065	0.068
One‐factor with error covariance		126.43 [89]	<.01	0.917	0.902	0.067	>.10	0.067
Pre‐registered mediation models using Bayes estimator
Primary State–State Mediation Model	Figure 2 – Panel A		(.48)	0.999	1.00	0.022	>.20
Adjusted Primary State–State Mediation Model	Figure 3 – Panel A		(.25)	0.960	0.893	0.105	>.20
Secondary Trait–State Mediation Model	Figure 2 – Panel B		(.50)	1.00	1.00	0.00	>.20
Adjusted Secondary Trait–State Mediation Model	Figure 3 – Panel B		(.41)	0.985	0.886	0.112	>.20
Secondary Trait–Trait Mediation Model	Figure 2 – Panel C		(.50)	1.00	1.00	0.00	>.20
Adjusted Secondary Trait–Trait Mediation Model	Figure 3 – Panel C		(.48)	1.00	1.00	0.00	>.20
Non‐registered moderation and mediation models
State–State Moderation Model	Figure Supp‐III Panel A		(.45)	0.986	0.892	0.087	>.20
Adjusted State–State Moderation Model	Figure Supp‐III Panel B		(.35)	0.971	0.856	0.111	>.20
Primary State–State Mediation Model Adjusted for Trait Shame	Figure Supp‐V Panel A		(.45)	0.996	0.963	0.064	>.20
Adjusted primary State–State Mediation Model Adjusted for Trait Shame	Figure Supp‐V Panel B		(.47)	0.990	0.959	0.06	>.20

Abbreviations: ACP‐T2, Aversive Cognitive Perseveration assessed at time 2; CFA, confirmatory factor analysis; CFI, comparative fit index; PPP, posterior predictive p‐value; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual; TLI, Tucker–Lewis Index.

^*The lower bound 90% confidence interval in all RSMEA indices with p > 0.10 reached 0.00.

Furthermore, as Table 2 reports, the One Factor model showed a slightly improved fit overall, though some indices were still borderline and valid factor loadings (range: 0.347–0.805). Upon inspection of modification indices, we introduced an error covariance between items RRS7 (“why I can't respond in a better way.”) and RRS5 (“why can't I handle things better?”), which contain highly similar wording. This modification resulted in a satisfactory fit, especially in terms of CFI = 0.917, TLI = 0.902, and RMSEA = 0.067 (see Table 2), with valid factor loadings from 0.344 to 0.781 (α and Ω = 0.91).

Moreover, we also tested the factor structure for trait ACP to ensure its construct validity. Primarily, we rescaled the composite scores to 0–100 scores since the original measures used inconsistent measurement scaling (see Supplementary Material‐Primary Measures). The CFA results demonstrated an acceptable one‐factor structure (fully identified), with factor loadings ranging from 0.638 to 0.688 (α and Ω = 0.70). Trait ACP correlated moderately with the reactive‐state ACP components (r = 0.53–0.62, p < .001) and showed a stronger association with the aggregated reactive‐state ACP score (r = 0.64, p < .001). Together, these findings further supported that ACP could be treated as a coherent one‐factor construct in this sample at both the trait and reactive‐state operationalizations.

Bivariate correlations

CEA history showed significantly positive associations with trait shame (r = 0.36, p < .001) and trait ACP (r = 0.29, p < .001), whereas it showed no significant associations with reactive‐state ACP‐T2, baseline‐state shame‐T0, reactive‐state shame‐T1 and reactive‐state engagement strategy use‐T2. The adjusted score of general maltreatment also showed the same pattern of results, except for a significantly positive association with baseline‐state shame‐T0 (r = 0.22, p < .05). Furthermore, trait shame showed moderate associations with baseline‐ and reactive‐state shame (r = 0.33 and 0.31, p < .001, respectively), whereas trait ACP showed a strong association with reactive‐state ACP‐T2 (r = 0.63, p < .001) and trait shame (r = 0.69, p < .001) (see Table 1 for further details).

Pre‐registered state mediation model

Table 2 provides the model fit, and Figure 2 reports the results for the primary mediation analysis that tested if the association of CEA history and reactive‐state ACP‐T2 was mediated by reactive‐state shame‐T1, controlling for baseline‐state shame‐T0 (Panel A). Contrary to expectations, reactive‐state shame‐T1 showed a negative and non‐significant indirect effect (β _ab  = −0.08, 95% CrI = [−0.20, 0.01]). However, there was a significantly positive association between reactive‐state shame‐T1 and ACP‐T2 (β _b  = 0.50, 95% CrI = [0.32, 0.65]). CEA history also showed non‐significant associations with reactive‐state shame‐T1 (β _a  = −0.17, 95% CrI = [−0.36, 0.02]) and ACP‐T2 (β _c’ = 0.12, 95% CrI = [−0.06, 0.29]). Moreover, when the model was adjusted by reactive‐state negative affect‐T1 and engagement strategy use‐T2, in addition to the adjusted score of general maltreatment, reactive‐state shame‐T1 still predicted greater ACP‐T2 (β _ab  = 0.36, 95% CrI = [0.12, 0.58]), while no notable alterations were observed regarding the direct and indirect effects of CEA history (Figure 3, Panel A).

FIGURE 2.

Estimated registered models. (A) Primary Reactive‐State Mediation Model. (B) Secondary Trait–State Mediation Model. (C) Secondary Trait–Trait Mediation Model.

FIGURE 3.

Adjusted estimated registered models. (A) Adjusted Primary Reactive‐State Mediation Model. (B) Adjusted Secondary Trait–State Mediation Model. (C) Adjusted Secondary Trait–Trait Mediation Model.

Pre‐registered trait mediation models

Table 1 provides the model fit, and Figure 2 reports the results for the secondary mediation models that tested whether trait shame mediated the association between CEA history and reactive‐state ACP‐T2 (Panel B) and trait ACP (Panel C). In terms of reactive‐state ACP‐T2, CEA history showed a non‐significant total effect (β _c  = 0.07, 95% CrI = [−0.14, 0.26]) and a non‐significant direct effect (β _c’ = −0.10, 95% CrI = [−0.29, 0.09]). However, consistent with hypotheses, trait shame showed a positive indirect effect (β _ab  = 0.16, 95% CrI = [0.07, 0.28]) in which CEA history was associated with greater trait shame (β _a  = 0.35, 95% CrI = [0.16, 0.51]), which in turn was associated with greater reactive‐state ACP‐T2 (β _b  = 0.49, 95% CrI = [0.30, 0.64]). The results held when controlling for reactive‐state engagement strategy use‐T2 and the adjusted score of general maltreatment (Figure 3, Panel B and C, respectively).

In terms of the model with trait shame and trait ACP (Figure 2, Panel C), CEA history showed a significantly positive total effect (β _c  = 0.28, 95% CrI = [0.08, 0.45]) that substantially decreased to a non‐significant direct effect (β _c’ = 0.04, 95% CrI = [−0.12, 0.20]). Consistent with hypotheses, a significantly positive indirect effect was observed (β _ab  = 0.23, 95% CrI = [0.11, 0.36]), in which CEA history was associated with greater trait shame (β _a  = 0.35, 95% CrI = [0.16, 0.51]), which in turn was associated with greater trait ACP (β _b  = 0.68, 95% CrI = [0.54, 0.79]). These results held when controlling for the adjusted score of general maltreatment (Figure 3, Panel C).

Non‐registered post hoc models

Despite a strong mediation effect through trait shame, CEA history showed non‐significant associations with state shame and state ACP in the primary state mediation model. We conducted a post‐hoc moderation analysis due to two points of concern. Empirical research has associated CEA history with bi‐phasic reactions to stress stimuli, involving hyper‐physiological followed by hypo‐physiological activations as stressful situations unfold (Duprey et al., 2021; Neuner, 2023; Rosenbaum et al., 2024). Additionally, reduced variability (e.g., ceiling effects) in reactive‐state ACP use in individuals with elevated CEA history (Amini‐Tehrani et al., 2025) could otherwise underlie the observed lack of associations. Thus, we examined whether this association holds across the levels of CEA history as a moderator.

Table 2 provides the model fit indices. As Figure Supp‐III depicts, controlling for baseline‐state shame‐T0, there was a significantly negative moderation effect for CEA history (β _interact = −0.49, 95% CrI = [−0.69, −0.09], Panel A), which was robust to the covariates (e.g., reactive‐state negative affect‐T1, engagement strategy use‐T2 and general maltreatment, β _interact = −0.49, 95% CrI = [−0.60, −0.21], Panel B).

Furthermore, the Johnson–Neyman plot indicated that the positive slope of reactive‐state shame‐T1 on reactive‐state ACP‐T2 weakened as CEA scores increased, and it became non‐significant at a CEA history score above 12 (n = 22 with CEA scores >12, Figure Supp‐IV), which denotes a moderate‐to‐severe CEA history (Bernstein et al., 2003). Thus, the link between reactive‐state shame‐T1 and ACP‐T2 was most evident for people with CEA history in the below‐moderate range.

The post‐hoc moderation results raised the question about the response pattern underlying the non‐significant association of reactive‐state shame and ACP use in participants with moderate‐to‐severe CEA history. Thus, to clarify the unique associations of CEA history in the current situation, we adjusted the primary reactive‐state mediation model (Figures 1 and 2, Panel A) by trait shame, due to its significant associations with all the primary variables (see Table 2 for the model fit and the Supplementary Material‐Figure Supp‐V). The results showed that the non‐significant association between CEA history and reactive‐state shame (β _a  = −0.17, CrI = [−0.36, 0.02], Figure 1‐Panel A) became significant (β _a  = −0.28, CrI = [−0.46, −0.08], Supplementary Material‐Figure Supp‐V, Panel A). Importantly, the nil indirect effect of CEA history on reactive‐state ACP use (β _c’ = −0.08, CrI = [−0.20, 0.01], Figure 1‐Panel A) also turned significant (β _c’ = −0.11, CrI = [−0.22, 0.03], Supplementary Material‐Figure Supp‐V, Panel A), suggesting a significant reduction in ACP use as CEA history increased.

These results held when the model was further adjusted for reactive‐state negative affect‐T1, engagement strategy use‐T2, and adjusted score of general maltreatment (Figure Supp‐V, Panel B, see Table 2 for the model fit). Overall, there was a tentative possibility that CEA history was associated with decreased reactive‐state shame (i.e., attenuated shame reactivity), followed by reduced cognitive processing, a pattern that was highlighted in individuals with moderate‐to‐severe CEA history in this sample.

Non‐registered sensitivity analysis

Given that empirical studies on reactive‐state ACP and trait ACP as a coherent construct were minimal, we retested the adjusted versions of reactive‐state and trait models, with ACP components as independent outcomes. All the retested models yielded comparable associations and significance levels, further supporting that the shared variance among ACP components may represent a coherent cognitive reactivity in the current situation. The detailed results are provided in the Supplementary Material‐Sensitivity Analysis.

DISCUSSION

Survivors of CEA are at particularly high risk of poor emotion regulation, especially overreliance on ACP, including brooding rumination, experiential avoidance and emotional non‐acceptance. Identifying the factors that help attenuate this relationship is essential for effective interventions. In this study, we tested the association of CEA history and ACP and whether shame mediates this association in reactive‐state and trait modes. Overall, some expectations were validated, in which ACP formed a broader strategy factor in a reactive‐state episode in the lab (McMahon & Naragon‐Gainey, 2019; Naragon‐Gainey et al., 2017), and trait response tendencies in shame and ACP use tended to manifest in a reactive‐state mode in the lab (Gratz et al., 2007). However, the findings partially supported the expected indirect effect, which emerged in trait shame but not state shame models. Thus, we first discuss the association of shame and ACP use and then the role of CEA history, suggesting avenues for future studies.

The associations between shame and ACP use in the trait and reactive‐state modes remained robust across models and covariates. This overarching finding primarily extends previous evidence that reactive‐state shame intensifies rumination (Cândea & Szentágotai‐Tătar, 2017; Milia et al., 2021; Vrshek‐Schallhorn et al., 2019) and further demonstrates that this process also encompasses experiential avoidance and non‐accepting meta‐emotions (Gratz & Roemer, 2004; Hayes et al., 1996; Mitmansgruber et al., 2009). The associations of reactive‐state shame and ACP use above and beyond negative affect in the lab suggest that the current emotion regulation responses were not solely driven by emotional reactivity tendencies. Consistent with previous evidence (e.g., trait associations: Cenkner et al., 2023; Leonard et al., 2020; Mirabile et al., 2023; and state associations: Vrshek‐Schallhorn et al., 2019), we suggest that the shame–ACP association reflects a broader negative self‐referential emotional processing. This response‐focused emotion regulation tendency tentatively describes that individuals who show greater shame sensitivity tend to exhibit higher levels of ACP use.

Notably, the shame‐ACP association potentially embodies an integrated tendency to activate and sustain shame‐induced negative self‐awareness (see Duval & Wicklund, 1972, for a detailed description of self‐awareness processing). One could speculate that such a tendency could represent a broader emotion regulation vulnerability (e.g., negative self‐focused attention, Pyszczynski & Greenberg, 1987), especially given the extensive associations of both trait shame (Laving et al., 2023; Sheehy et al., 2019; Shi et al., 2021) and ACP use (Aldao et al., 2010; Hayes et al., 1996; Naragon‐Gainey et al., 2017) with internalizing symptoms.

The role of CEA history

Pre‐registered findings

The pre‐registered models broadly investigated shame as a mediator of the association of CEA and ACP use. The current mediation models broadly supported trait shame to mediate the effect of CEA history on trait and reactive‐state ACP use, while the evidence for increased reactive‐state shame was mixed. Additionally, the indirect effect of CEA history on reactive‐state ACP use through trait shame, particularly despite the tentative absence of the total effect, might indicate that CEA history may function as one notable developmental factor, among others such as cultural and genetic factors, which contribute to trait shame and its subsequent functions (Kouri et al., 2023; Ross et al., 2019). Overall, the indirect effects via trait shame support that CEA history could contribute to response tendencies embodied in the shame–ACP association, independent of the effect of physical abuse, physical neglect and emotional neglect histories.

However, the nil associations between CEA history and reactive‐state shame and ACP use in the primary mediation model contradict our expectations based on previous associations of CEA history and stress reactivity in daily life (Shapero et al., 2014). To account for this, it is plausible that the current self‐esteem threatening stimulus using a direct message of removal could induce relatively higher levels of shame, as compared to stimuli using anticipated evaluation (Zoccola et al., 2012) or self‐shame imagery paradigm (Kouri et al., 2023). Consequently, the individual differences associated with CEA history might have been overridden by participants' reactive‐state responses that reflect their ‘capacity to experience shame and ACP use,’ regardless of the frequency of associated experiences in real life (i.e., traits) (Dang et al., 2020). On the other hand, the current shame induction did reveal the expected shame–ACP association, identifying individual differences in shame and ACP responses in the lab situation. Thus, one possible explanation for our observation regarding the decoupled trait versus state response tendencies could be attributed to potential individual differences in the strength of the shame–ACP association across the levels of CEA history. The remaining text draws on this suggestion, providing directions for future studies.

Post hoc non‐registered findings

The post‐hoc analyses could tentatively illustrate the inconsistencies between trait versus reactive‐state shame–ACP association with respect to the role of CEA history. Specifically, the post‐hoc analyses highlighted the possibility that the observed response patterns were influenced by the attenuated stress reactivity. In fact, studies on acute stress reactivity have frequently observed hypo‐physiological reactivity in terms of reduced heart rate variability, skin conductance, cortisol level and hypothalamic–pituitary–adrenal (HPA) axis in individuals with higher CEA history (Duprey et al., 2021; Rosenbaum et al., 2024) and with trait perseverative negative thinking (PNT) (Ottaviani et al., 2016). Importantly, most of these studies concomitantly induced shame and stress in the lab (e.g., Trier Social Stress Test) (Allen et al., 2017).

Thus, the current response pattern could raise the possibility of attenuated shame reactivity, a response pattern similar to attenuated stress reactivity, in which shame–ACP association might be contingent on other CEA‐associated factors involved in difficulties with emotion generation and regulation. Due to their tentative suggestions, future research should consider the role of CEA further and explore a possibility of an integrative model of state emotion dysregulation in self‐esteem threatening situations.

Clinical implications

The current study sheds light on the association of CEA history with ACP use as one type of response‐focused emotion regulation (Weissman et al., 2019). The association of shame and ACP may contribute to symptom formation due to reactive use in response to stressful stimuli (Pyszczynski & Greenberg, 1987). Thus, we encourage research and clinical practice to give deliberate attention to shame‐based self‐representations underlying ACP use associated with CEA history.

This suggestion particularly follows Lewis's advice (Lewis, 1987) and Greenberg's illustration (Greenberg, 2015): it is necessary for clients to, very early in therapy, distinguish between shame and general distress (e.g., feeling ill or upset) (Lewis, 1987, p. 21). This instruction seems fundamental given that clients may “bypass shame” by evading its pain through off‐topic elaborations and daily ruminations (Eterović, 2020; Lewis, 1987). Given that shame inherently encourages avoidance and persists through multiple personal and social processes (Leeming & Boyle, 2004), mental health practitioners can actively explore deeply‐held self‐representations and the associated ACP, especially in individuals with elevated CEA history.

Strengths and limitations

Several factors speak to the study's strengths and the reliability of our findings. Specifically, owing to the induction narrative, the current study set apart the real‐time induction of shame and the subsequent emotion regulation process, making the spontaneous use of both aversive cognitive reactions and potential reappraisal observable (Gross, 2015). Moreover, research has shown that state emotional reactivity tends to be potentiated even in the presence of possible “hypothesis‐guessing” (Linden et al., 2010), which could support that our observations remained robust to potential demand characteristics of the shame induction task (e.g., deliberate self‐reflection). We also applied manipulation checks across several time points (in‐experiment questions, attention check, debriefing) and used inattentive response checks, which increased confidence in the internal validity of the shame induction and self‐reports.

However, due to several internal and external validity concerns, our findings should be interpreted with caution. Importantly, we terminated data collection on the pre‐registered time (November 2024), in addition to resource and time constraints. Thus, the current underpowered sample size may limit the estimation for the state models, rendering especially the non‐significant associations tentative suggestions, rather than a true lack of individual differences across the levels of CEA history. Additionally, the current results relied entirely on self‐reported trait and state measures collected within a single session, which may contribute to common method bias and limit the validity of the shared variance observed in the mediation models.

Furthermore, the current sample was drawn from one university campus. Thus, despite a distribution of CEA history that aligns with population‐level estimates (Mathews et al., 2023), generalizability to community populations is limited. Importantly, generalizability to clinical samples was further constrained by the exclusion criterion of severe levels of depressive symptoms on DASS‐21 (Lovibond & Lovibond, 1995), though the current sample did include a large representation of mild and moderate depression symptoms (>50% of participants). In this regard, research has highlighted attenuated emotional reactivity in patients with major depression disorder (Bylsma et al., 2008), which has nuanced implications in our observations. Specifically, it is plausible to argue that this specific exclusion criterion could minimize the risk of attenuated emotional reactivity due to depressive symptoms. However, our inference as to potential attenuated shame reactivity was further restricted by reliance on subjective emotional experience, which could be better evidenced by interoceptive awareness and physiological markers of emotional reactivity (Beauchaine, 2012; Bylsma et al., 2008).

Moreover, the current findings and suggestions may not generalize to CEA history in the context of sexual abuse history. Empirical research suggested that childhood sexual abuse was associated with rumination underlying maladaptive behaviours (Sarin & Nolen‐Hoeksema, 2010). However, CEA history was longitudinally associated with higher distress through daily avoidant coping, while sexual abuse history showed no longitudinal associations with emotional avoidant tendencies in daily life (Nguyen‐Feng et al., 2017). Overall, sexual abuse history may form a unique traumatic context among maltreatment types (Kim et al., 2023; MacGinley et al., 2019). Thus, approaches that consider combined maltreatment history (Rivera et al., 2018) could best tap into the broader maltreatment context, especially CEA in the context of sexual abuse history.

Moreover, the current shame induction was contingent on multi‐dimensional cues, which could concomitantly induce some levels of acute shock (Turner, 2014) and fear (Fink‐Lamotte et al., 2023), in addition to goal striving efforts associated with CEA history in competitive tasks (Upenieks & Ford‐Robertson, 2023). As a result, it is plausible that this emotionally convoluted situation might trigger response tendencies associated with stress reactivity (Schär et al., 2022), apart from ACP thought patterns. In parallel, the observed ACP thought patterns also could be partly driven by these other potentiated affects that were not controlled for, highlighting that individual differences in ACP use may not be exclusive to shame‐based self‐representations.

Finally, theoreticians have phenomenologically distinguished shame tendencies from other negative self‐conscious feelings, for example embarrassment, humiliation and being disgusted (Lewis, 1987, 2019; Tangney et al., 1996; Tracy & Robins, 2006). Strategically, rather than the global shame concept (e.g., ‘feeling ashamed,’ Mackinnon et al., 1999), our shame operationalization was based on negative self‐representations that carry increasingly stronger affective tones (e.g., feeling ‘stupid’ versus ‘humiliated’) (Harder & Zalma, 1990). Given the diversity in shame measurement and a lack of coherence between shame conceptualization and operationalization in the existing literature (Garofalo et al., 2025), the current observations may be exclusive to the adopted operationalization, limiting replicability and generalizability, especially with respect to sources of shame other than CEA history.

Future directions

Overall, to address the emotion regulation difficulties associated with CEA history in adulthood, further research is needed to address limitations and elucidate the involved mechanisms. Primarily, future studies should further test the post‐hoc moderation hypothesis we conducted to see if subjective experience of shame attenuates immediately following the induction. The role of situation in the manifestation of dispositions associated with CEA history may also lie in the trajectories of the shame–ACP association across online (in‐session) and offline (post‐session) time periods. Our study specifically tested in‐session reactive‐state tendencies, leaving post‐session state tendencies for future studies. Specifically, subsequent state post‐session negative self‐referential processing (e.g., intrusions, deliberate ruminations) (Zoccola et al., 2012) might reinforce a core negative self‐representation (e.g., ‘I am a failure’), potentially forming self‐congruent memory of shame.

Additionally, future studies could integrate stress reactivity and response‐focused emotion regulation to investigate microprocesses as shame induction progresses. As Neuner (2023) summarized, childhood maltreatment and specifically CEA history has been associated with patterns of bi‐phasic stress reactivity, in which physiological hyperactivation is followed by abruptly hypoactivation in response to stress. Thus, a repeated measures design that assembles in‐session and post‐session state experiences could trace individuals' mental processing (e.g., shame, self‐referential thought patterns, state emotion regulation ability use) across multiple timepoints, together with the concomitant physiological indicators (e.g., skin conductance) (England‐Mason et al., 2017). Such triangulation of pre‐, post‐ and later assessments could shed light on whether there might be a state of reactive emotion dysregulation in which acutely reduced mental representations in response to internal negative self‐states (instantiated in reduced reactive‐state shame and ACP use in the lab) are coupled with physiological hypoactivation in response to external distressing stimuli.

In terms of the underlying processes, previous studies based on total childhood maltreatment history have attributed such aberrant response tendencies in the lab to trait emotion dysregulation (England‐Mason et al., 2017), trait detachment dissociation (Seitz et al., 2024) and trait dismissive attachment tendencies (White et al., 2012). One possible account of this is that such presentations stemmed from CEA‐associated developmental adaptations in various neural substrates associated with self‐awareness and self‐referential processing (Teicher et al., 2014; van der Werff et al., 2013), emotional processing (Gerin et al., 2023), in addition to amygdala reactivity (Hein & Monk, 2017). These multi‐modal functional adaptations could eventually strengthen the child's avoidant capacity in a chronically harsh environment. In adulthood, given that shame‐based self‐representations still carry intense affective components, an aberrant multi‐modal emotion regulation might acutely help over‐modulate subjective distress (e.g., Ji et al., 2024). Further research is required to fully elucidate the specifics of emotion generation and processing associated with CEA history, especially in situations involving self‐esteem threatening cues.

Furthermore, although the results tentatively supported that shame reaction appeared in increasingly harsher self‐representations in this sample to account for ACP use in the lab, further research is needed to justify whether the shame–ACP association is a coherent self‐processing phenomenon. Specifically, our operationalization of state shame allowed for broader affective self‐categorization tendencies in shame situations (Barrett, 2016; Harder et al., 1993; Tracy & Robins, 2006). However, future studies can test whether the individual differences in ACP use were better explained by specific appraisals underlying reactive‐state shame. That is, the shame–ACP association and its potential attenuated form may not be exclusive to a sense of shame that strongly focuses on flawed‐self appraisals (Lewis, 1987). Alternatively, ACP use may arise in shame situations due to fear appraisals that concern a sense of physical threat, guilt appraisals that are limited to particular behaviours and reparation, embarrassment appraisals that focus on a singular situation and humiliation appraisals that focus on unfair treatment (Fink‐Lamotte et al., 2023; Gilbert, 1998; Lewis, 2019; Tracy & Robins, 2006). In the same vein, shame operationalization could be further distinguished in terms of external and internal shame (Kim et al., 2011), and their respective self‐representations could be investigated in terms of their specific self‐attributional appraisals (e.g., in the view of oneself – ‘I screwed up’ – or others – ‘they hate me now’). Thus, future studies may investigate the role of CEA history in individual differences in perceptions and appraisals of self‐esteem threatening situations to explore the contingencies of affective self‐categorization that encourages subsequent ACP use (Barrett, 2016).

CONCLUSION

CEA history may shape trait shame tendencies in ways that prioritize ACP use as a response‐focused emotion regulation tendency. However, despite elevated ACP thought patterns, reactive‐state shame did not manifest in individuals with elevated CEA history. Although the associated ACP use was reduced in the lab, it nevertheless habitually operated in response to self‐referent stimuli. Further research is required to explore the processes underlying ACP use and the potential involvement of aberrant stress reactivity in modulating shame reactions associated with higher CEA history.

AUTHOR CONTRIBUTIONS

Mohammadali Amini‐Tehrani: Conceptualization; investigation; writing – original draft; methodology; validation; visualization; formal analysis; project administration; data curation; writing – review and editing. Michael Weinborn: Supervision; writing – review and editing; methodology; project administration. Kenneth Sim: Investigation; writing – review and editing. Jeneva L. Ohan: Conceptualization; supervision; writing – review and editing. Kristin Naragon‐Gainey: Conceptualization; project administration; supervision; writing – review and editing; methodology; formal analysis.

FUNDING INFORMATION

This research did not receive any specific grant from funding agencies in the public, commercial or not‐for‐profit sectors.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Data S1: Supplementary Methods and Materials.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author [MAT] upon reasonable request.