Perceptions of Physical, Sexual, and Psychological Violence Stories: A Registered Report

Interpersonal violence (IPV) refers to abuse committed within an ongoing relationship. Such abuse can be sexual, physical, or psychological. Psychological abuse, the least-studied form of IPV (Jewkes, 2010; Mountjoy et al., 2016), consists of a pattern of deliberate, non-physically aggressive relationship behaviors that are emotionally or mentally harmful, controlling, and/or threatening (Breiding et al., 2015). These acts may be overt (e.g., name-calling, humiliation, threats of physical or sexual violence, expressing anger in a way that seems dangerous) or covert and manipulative (e.g., exploitation of a victim or perpetrator’s vulnerability such as immigration status, or excessive monitoring of a partner’s whereabouts or personal communication; Breiding et al., 2015; Doherty et al., 2008).

We hypothesize that relative to the other types of IPV, psychological abuse stories and their storytellers will be least-favorably received by audiences, for several reasons. First, unlike the discrete events that comprise acts of sexual and physical violence, psychological abuse is more diffuse. Psychological abuse represents an accumulation of controlling, degrading actions over time. Any single incident of psychological aggression, such as a single act of name-calling, does not have the narrative weight of a single incident of rape or physical assault. As such, psychological abuse may be harder for survivors to succinctly and impactfully narrate in a “story,” failing to evoke the cumulative emotional weight of the experience for audiences.

Second, although all types of IPV within ongoing relationships tend to happen behind closed doors, physical and sexual violence have the potential to leave objective physical, and at times, visible, evidence. Without leaving objective and/or visible evidence, psychological abuse may be viewed as less credible (Candela, 2016). This perception is belied by evidence that psychological abuse can have an even more severe impact than physical or sexual abuse in the long term (Dutton et al., 2006; Follingstad et al., 1990). Ambiguity creates more leeway for bystanders to dismiss psychologically abusive events as a misunderstanding or victim overreaction, or to minimize the severity of the events. As such, when a psychological (versus sexual or physical) abuse survivor tells their story, simply explaining the events may expose these survivors to more negative personal judgments.

One way that survivors of various forms of trauma evoke audience support and acceptance is to redeem their stories. In a redemptive narrative, the storyteller overcomes initial adversity by attaining personal growth, liberation, or recovery. Redemption projects personal strength and resilience, fitting within a culturally valued narrative framework (McAdams, 2006; McLean et al., 2020: McLean & Syed, 2015). In an extension of Delker et al. (2020), however, we anticipate that redemption will be less available to survivors of psychological violence compared to survivors of sexual and physical violence. In particular, we predict that the ability of redemption to boost storyteller likeability will be overshadowed by stigmatizing responses to psychological (relative to sexual and physical) abuse survivors. That is, audiences will respond to psychological (versus sexual and physical) abuse storytellers with less empathic concern, greater desire for personal distance, and more negative attributions about their personalities, even when stories end redemptively.

Additionally, although we anticipate that differences among the three types of IPV will shape varying reactions to the three types of stories, all types of IPV share a core commonality: they are more stigmatizing than traumas caused naturally (e.g., natural disaster) or by people accidentally (e.g., vehicle collision). That is, when survivors of IPV disclose their traumatic experiences in personal and professional settings, they tend to be met with more negative reactions such as disbelief and victim-blame (Dworkin et al., 2019; Kennedy & Prock, 2018), though this research has tended to focus on sexual and physical, rather than psychological, violence. As such, we anticipate that, when compared en masse to non-IPV traumas, all types of IPV traumas and their storytellers will be perceived less favorably, replicating and extending Delker et al. (2020).

In this vignette-based, experimental Registered Report study, participants were randomly assigned to read and evaluate a narrative from a set of first-person trauma narratives which varied along two dimensions: type of trauma story (IPV vs. non-IPV) and ending (redemptive vs. negative). This study was a conceptual (rather than direct) replication and extension of McLean et al. (2020) and Delker et al. (2020), across several dimensions. First, this study examined audience perceptions of three major types of IPV, whereas the previous studies focused on sexual violence only (in comparison to non-IPV traumas). Second, whereas the story of adult sexual violence in the previous studies depicted a case of acquaintance rape, the present study depicted cases of sexual, psychological, and physical violence in the context of ongoing relationships. As all three IPV stories occur in adult relationships, we also changed the chronic illness non-IPV story from a child to adult protagonist for this study. Finally, in this study, to assess audience perceptions more comprehensively, we added measures of empathic concern and stigmatizing attitudes toward the storytellers.

As in the prior studies, the gender and other demographic characteristics of the survivor-storytellers were not explicitly stated, though audiences may infer that the storytellers were female based on gendered assumptions about bodies and relationships. For instance, in the chronic illness story, the narrator describes experiencing breast cancer, and in the IPV stories, the presumptively female narrator uses he/him/his pronouns to describe the actions of the unnamed perpetrator. This male-female perpetrator-victim pairing in the IPV stories reflects broader societal patterns and gendered inequalities in which the majority of IPV perpetrators are men, and the majority of IPV victims are women (Black et al., 2011). In the case of sexual violence, for instance, approximately 1 in 5 women (18.3%, or 22 million women) versus 1 in 71 men (1.4%, or 1.6 million men) report experiencing rape in their lifetime (Black et al., 2011). The IPV story protagonists in this study are gendered accordingly, as an empirical starting point that reflects the most common, numerous populations of who is victimized (and who victimizes others) across the Americas (Bott et al., 2019) and around the world (World Health Organization, 2018). However, we want to wholeheartedly acknowledge the reality of relationship-based violence faced by cisgender men and sexual and gender minority populations (Coulter et al., 2017; McCauley et al., 2018), and to invite further research that can speak to the additional layers of cultural stigma that these survivors face in telling their stories.

The first set of hypotheses compare reactions to IPV versus non-IPV trauma stories.

Hypothesis 1a. We predicted that, compared to non-IPV stories with negative endings, IPV stories with negative endings would be perceived as more difficult to tell and less likely to be told; their storytellers as less likeable and having less positive, socially desirable personality traits; and audience responses would involve less empathy and more stigmatizing attitudes (including victim-blame).

Hypothesis 1b. We predicted that narrative redemption would moderate the effect of trauma story type (IPV vs. non-IPV) on story perceptions, such that redemptive story endings would enhance the positivity of audience perceptions of non-IPV trauma stories only (vs. IPV traumas).

The second set of hypotheses compare reactions to the three types of IPV.

Hypothesis 2a. Compared to sexual or physical violence stories with negative endings, we predicted that psychological violence stories with negative endings would be perceived as more difficult to tell and less likely to be told; their storytellers as less likeable and having less positive personality traits; and audience responses would involve less empathy and more stigmatizing attitudes (including victim-blame).

Hypothesis 2b. We predicted that narrative redemption would moderate the effect of IPV trauma type (sexual, physical, psychological) on story perceptions, such that redemptive story endings would enhance the positivity of audience perceptions of sexual and physical violence stories to a greater degree than psychological violence stories. However, consistent with Hypothesis 1b, we anticipated that overall positivity of IPV story perceptions, even with redemptive endings, would not exceed that of non-IPV stories with redemptive endings

For both hypotheses 2a and 2b, we left as exploratory whether there would be differences between physical and sexual violence in perception of stories.

The covariates for the analyses in this study were the same as the covariates in McLean et al. (2020) and Delker et al. (2020): personal characteristics connected to trauma exposure, specifically participant gender and participant trauma history. We included these covariates in our analyses due to the possibility that greater personal experience with and/or lifetime risk for the IPV traumas being described might influence perceptions of the storytellers.

We administered this study online to adult participants in the United States. As a Registered Report, the proposed study methodology and analysis were pre-registered as a Registered Report Protocol on the Open Science Framework (OSF) following Stage 1 peer review and in-principle acceptance from this journal, and prior to data collection (OSF Link: https://osf.io/z76ue/). As such, the study rationale, research questions, hypotheses, and proposed study protocol described below were peer-reviewed prior to any research being conducted.

Three samples (N = 1,045) consisted of U.S. adults recruited via a Northwestern public university student participant pool (University Sample n = 360) and via Prolific, a crowdsourced online labor platform, census-matched for age, race, and gender (Nationally Representative Sample 1, n = 342 and Sample 2, n = 343). A descriptive summary of participant demographic characteristics for each sample can be found in Table 1. Overall, the gender and race characteristics of the two Nationally Representative samples reflect U.S. census data (U.S. Census Bureau, 2021), with the majority (about three-quarters of the total sample) self-identifying as white. The University sample was majority (about 65%) female and more diverse in terms of gender expansiveness and sexual identity, with more than 40% of the sample not identifying as heterosexual. The number of Latinx individuals in this study was consistent with U.S. census data for one nationally representative sample (19%), but lower than census estimates for the other two samples.

Based on anticipated effect sizes (medium effect sizes for Hypotheses 1a and 1b and small-to-medium for 2a and 2b), alpha at .05, and a priori power at .90, the planned sample size for this study was N = 342 (n = 30 participants per 12 conditions), per each of the three samples.

Due to a programming error in the study randomization process discovered after data acquisition was complete, we did not reach the planned n per condition across all 12 experimental conditions in the initial round of data collection. Although participants were randomly assigned to experimental conditions, the randomization did not occur evenly. Due to research funding limits, we could not collect additional data to reach the planned n = 30 per condition in Nationally Representative Sample 1 (n_range = 20-34) nor 2 (n_range = 22-33). We were able to collect additional data in the University setting to reach the planned n = 30 per condition in that sample. More information documenting the process of additional data collection can be found in our Laboratory Log on OSF. In the Analysis Plan section below, we explain how we addressed statistical power for the planned analyses.

Inclusion criteria were U.S. adults age 18 and over with English proficiency to complete the English-language surveys. Additionally, we proposed that if participants did not pass our validity checks, their data would be excluded from the analysis and replaced with the data of additional participants who passed the validity checks (complete study in one sitting, do not complete study too quickly, i.e., 3 or more standard deviations below the mean completion time). Consistent with prior vignette-based online research by our team (McLean et al., 2020) and with national estimates of careless responding in survey-based online research (Meade & Craig, 2012), we anticipated needing to replace the data of approximately 12% of our starting N.

After data acquisition, we discovered that while the mean time to study completion was consistent with our expectations, the standard deviation was such that operationalizing “too quick” completion as three or more standard deviations below the mean completion time would be a negative number. As such, we were not able to exclude cases from the data based on that criterion, though we did exclude the data of cases who did not complete the study in one sitting (n = 10 across the three samples). Instead of using the metric of standard deviation for time to complete, we excluded those who completed the study in under 3 minutes (n = 8 in the University Sample; none were under 3 minutes in the Prolific samples). Three minutes was chosen post-hoc based on what the research team determined, after trial-running the study, to be an excessively quick reading time.

Our university’s Research Compliance Organization approved the study protocol (#4369EX21). Participants completed the study on Qualtrics.com. Individuals in all samples signed up to participate in the study titled “Evaluating Stories.” Participants from Prolific received financial incentive upon valid completion of the surveys. University students were recruited through a university-based online research management system and received course credit for study participation (alternative avenues for course credit were available). All participants provided informed consent electronically when they agreed to participate.

This approximately 10-minute study used a 6 (trauma story: IPV trauma [k = 3], non-IPV trauma [k = 3]) x 2 (ending: negative, redemptive) design. The three IPV trauma stories were sexual, physical, and psychological violence. The three non-IPV trauma stories were vehicle collision, natural disaster, and life-threatening illness (cancer). Participants were randomly assigned to read 1 of 12 (6 x 2) stories and then complete the study questionnaires.

Each of the 12 narratives described a traumatic experience from the first-person perspective of an anonymous author (see Text S1). Narratives were written with a consistent level of detail, in a similar style, in a conversational tone, and at a 5th-grade reading level (Flesch-Kincaid Grade Level = 5.45). Narratives were approximately 225 words each.

We designed the IPV narratives based on research and theory on components of IPV (Breiding et al., 2015; Herman, 2015; Snyder, 2019; Strauchler et al., 2004), along with public survivor stories from advocacy organizations such as the Rape, Abuse and Incest National Network (RAINN, 2020). We adapted the non-IPV narratives (car accident, natural disaster, life-threatening illness) from McLean et al. (2020) and Delker et al. (2020), expanding their length and emotional intensity to match that of the IPV narratives. As the IPV storytellers are adults, we also changed the life-threatening illness storyteller from a child to an adult.

The two possible trauma story-ending types were negative and redemptive. These endings manipulated whether the storyteller was still negatively impacted by their trauma, or if they had experienced personal redemption. One negative and one redemptive ending was used for all three IPV traumas, whereas non-IPV events were distinct enough that negative and redemptive endings were uniquely tailored to each.

In Stage 1 of this Registered Report, we presented the results of a pilot study testing the 12 narratives adapted for this study. The pilot study with the 12 vignettes served as a manipulation check that trauma stories with negative (vs. redemptive) endings were perceived to end significantly more negatively, for both IPV (d = 3.54) and non-IPV (d =1.20) traumas. In this Stage 2 Registered Report, pilot study results can be found in the Supplementary Materials (see Text S2).

A full list of scales, items, and response options for assessing participant perceptions of trauma stories and storytellers can be found in Table S1. Items were presented to participants in a fixed order.

Trauma history was indexed as number of types of traumatic events that participants self-report having experienced and/or witnessed. All items on the 17-item Life Events Checklist (LEC-5) that participants selected Happened to me and/or Witnessed it were summed to yield a traumatic event history total score for each participant (range=0-17; LEC-5). LEC-5 items represent events defined as traumatic by the Diagnostic and Statistical Manual of Mental Disorders (DSM-5; American Psychiatric Association, 2013).

Analyses were conducted separately by sample, using the software program SPSS (v28.0; IBM Corp, 2021). To test the study moderation hypotheses (1b and 2b), we employed the Hayes PROCESS macro (v4.0) for SPSS program to estimate bootstrapped indirect effects (Hayes, 2018).

For replication purposes, we proposed in our Stage 1 Registered Report to conduct a meta-analytic summary of the statistical results representing tests of our hypotheses across the three samples, using the R package metafor (v.2.1, Viechtbauer, 2010). In the time that elapsed between Stage 1 and the completion of study data collection and data cleaning, we determined that a mega-analytic approach would be preferable for methodological and practical reasons. First, given that we have the raw data available, a mega-analytic approach is preferred over a meta-analytic approach that estimates a population-level effect size from sample-level effects (Costafreda, 2009; Gonzalez et al., 2021). Second, with a mega-analytic approach that pools the raw data into a single sample and controls for the source of the data (University Sample 1, Nationally Representative Sample 1, Nationally Representative Sample 2) in our statistical models, we could generate more stable estimates of effect size than if we used the underpowered (due to uneven randomization) single samples.

To streamline the presentation of results in this registered report, we summarize the statistical findings within each of the three samples in the Supplementary Materials (see Text S3), and we summarize the pooled mega-analytic results below.

All data cleaning and coding steps undertaken below were pre-registered in our Stage 1 Registered Report. Dependent variables for each pre-registered statistical model below were indices of perceived difficulty of telling, perceived likelihood of telling, storyteller likeability, storyteller positive personality traits (operationalized as high Extraversion, Agreeableness, Conscientiousness, and Openness, and low Neuroticism), empathy toward storyteller, and stigmatizing attitudes (Y₁-Y₆). In our revised Stage 1 Registered Report, we proposed measuring participant victim-blame of the storyteller (Y₇), but due to a data collection error, this survey was not included in the final version of the study.

For each dependent variable, we combined the corresponding item variables into a single composite variable if Cronbach’s α >.60 for the corresponding items, or if r > .5 for the inter-item correlations of 2-item indices. All reliability statistics met the above criteria for acceptability, with the exception of the storyteller likeability items (n_items = 5) in the University Sample (α = .56) and the perceived likelihood of telling items (n_items = 2) in all 3 samples (r = .33, .38, .35). Additionally, the correlation of the perceived difficulty of telling items (n = 2), was near to being below the proposed cutoff in each of the 3 samples (r = .60, .58, .58). To address the poor internal consistency reliability of the above scales, we adjusted our data analysis approach in the following ways: (1) We dropped storyteller likeability as a dependent variable in the University Sample analyses (see Text S3); (2) we dropped University Sample participants from the mega-analytic test of our hypothesis about storyteller likeability; and (3) instead of using 2-item indices of perceived difficulty of telling and perceived likelihood of telling, we indexed each dependent variable with 1 item only: It would be emotionally difficult for the author to share their story (perceived difficulty) and The author is likely to share this story with others (perceived likelihood), each measured on a 5-point Likert-type scale.

Prior to conducting inferential tests of our hypotheses, data were inspected for normality and outliers. No extreme outlying values >|3| SD from the mean were present in the data, and no variable distributions were significantly skewed or kurtotic; thus, per our pre-registered plan, no variables needed to be transformed.

Covariates for all statistical models below were participant gender (C₁) and self-reported trauma history (C₂). For statistical analysis, participant gender was dummy-coded along a binary of 1 (man) and 0 (all gender identities ≠ man). Two additional covariates were added to the models for the mega-analysis to control for Sample (k = 3). The first dummy variable (C₃) was coded to compare the effect of Sample 1 (University Sample) versus Samples 2 and 3 (the Nationally Representative samples). The second dummy variable (C₄) was coded to compare the effect of Samples 2 versus 3 (the two Nationally Representative samples).

We specified a factorial ANOVA model with one main factor (trauma type: X₁ = IPV stories with negative endings; X₂ = non-IPV stories with negative endings), C_1,C_2,C_3, and C_4, predicting Y₁-Y₆.

We specified a moderation model representing the interaction between the two dummy-coded categorical variables, trauma type (X₁ =I PV; X₂ = non-IPV) and story ending (W₁= redemptive; W₂= negative), in predicting Y₁-Y₆, controlling for C_1,C_2,C_3, and C₄.

We specified a factorial ANOVA model with one main factor, IPV trauma type (X₁ = psychological violence story with negative ending; X₂ = sexual violence story with negative ending; X₃ = physical violence story with negative ending), C_1,C_2,C_3, and C₄ predicting Y₁-Y₆. In the factorial ANOVA model, IPV trauma type (X) will be contrast-coded as follows: psychological violence (X₁) versus sexual and physical violence (X_2-3); sexual violence (X₂) versus physical violence (X₃).

We specified a moderation model representing the interaction between the two dummy-coded categorical variables, IPV trauma type (X₁ = psychological; X₂ = sexual; X₃ = physical) and story ending (W₁= redemptive; W₂= negative), in predicting Y₁-Y₆, controlling for C_1,C_2,C_3, and C₄. Consistent with Hypothesis 2a, the predictor variable, IPV trauma type, will be dummy-coded in PROCESS as follows: psychological violence (X₁) versus sexual and physical violence (X_2-3); sexual violence (X₂) versus physical violence (X₃).

To estimate the magnitude of effects of our focal predictors in the factorial ANOVA models, we used partial eta-squared (η²_p), a commonly used measure of effect size for models with more than one independent variable (Fritz & Morris, 2018). Partial eta-squared describes the “proportion of variability associated with an effect when variability associated with all other effects is excluded from consideration” (Fritz & Morris, 2018, p. 607). In keeping with Cohen’s guidelines for psychological research, η²_p = .01 is considered a small effect, η²_p = .06 is considered medium, and η²_p = .14 is considered large.

The experimental stimuli in this study consisted of brief written personal accounts, about the length of several Tweets. We expect our results to generalize to situations in which participants read or view similar personal accounts of trauma by survivors. We also anticipate that results will be reproducible if the exact phrasing, content, and stylistic choices of the written trauma narratives vary, as long as the stories are consistent with psychological definitions of what makes events traumatic. The degree to which audience reactions to survivor stories may vary based on storytelling medium (e.g., written vs. audiovisual) or physical proximity of the storyteller (e.g., online or in-person; Fields et al., 2020) are empirical questions that can be addressed with conceptual replication studies.

As stated in the introduction, the survivor-storytellers in this study were gendered as women and the perpetrators in the IPV stories as men, presenting a potential limit to the generalizability of findings to the case of traumas experienced by women (and IPV traumas with male perpetrators and female victims). We felt that our experimental design would be obstructively complex if, in addition to experimentally manipulating trauma type and story ending, we had varied the genders and sexual identities of the protagonists. This would expand a 6x2 design with 12 conditions to a (at minimum) 6x2x2x2 design with 48 conditions. However, replication research can proceed incrementally, accounting for how additional layers of cultural stigma such as heteronormativity, homophobia, biphobia, transphobia, and cisgendernormativity might influence perceptions of survivor IPV (vs. non-IPV) stories.

The target population for our study was adults in the United States. We anticipate that a replication with any sample of adults in the United States would constitute a direct test of the same effects.

We have no reason to believe that the results depend on other characteristics of the participants, materials, or context.

Descriptive statistics for participant perceptions of trauma stories by trauma type, IPV type, and story ending can be found in Table 2.

We hypothesized that IPV stories with negative endings (versus non-IPV stories with negative endings) would be perceived as more difficult to share and less likely to be shared; their storytellers as less likeable and having fewer positive personality traits; and participant responses would be less empathetic and higher in stigmatizing attitudes towards the storyteller. Factorial ANOVA results for the test of this hypothesis in the combined sample can be found in Table 3.

Our hypothesis was supported for perceived difficulty sharing (η²_p = .068), likelihood of sharing (η²_p = .201), and storyteller positive personality traits (η²_p = .043). Participants rated IPV (versus non-IPV) stories with negative endings as more difficult to share, less likely to be shared, and the storytellers as having fewer positive personality traits.

There were no statistically significant differences between IPV and non-IPV stories with negative endings in perceived storyteller likeability and in participant empathy for the storyteller. Contrary to our hypothesis for stigmatizing attitudes toward the storyteller, participants reported less stigmatizing attitudes toward IPV (versus non-IPV) storytellers with negative endings (η²_p = .015).

Sample was a statistically significant predictor of perceived likelihood of sharing and empathy for the storyteller. Specifically, the Nationally Representative samples rated the storyteller as being significantly more likely to share their story (η²_p = .044) and endorsed significantly more empathy toward the storyteller (η²_p = .038) than did the University sample. Participants’ own self-reported history of traumatic life events was not a statistically significant predictor of any storyteller ratings or story perceptions. Male-identified participants reported significantly less perceived difficulty in telling the trauma story (η²_p = .014) and significantly less empathy for the storyteller (η²_p = .025) than those identifying as women, non-binary, or gender-queer, regardless of trauma story type.

We hypothesized that narrative redemption would moderate the effect of trauma story type (IPV vs. non-IPV) on story perceptions, such that redemptive story endings would enhance the positivity of audience perceptions of non-IPV trauma stories only (vs. IPV traumas).

Our moderation hypothesis was partially supported for perceived difficulty sharing (see Table 4 and Figure 1a). Redemptive story endings reduced the perceived difficulty of sharing non-IPV stories more than they reduced the perceived difficulty of sharing IPV stories. There was also a significant effect of sample, such that participants in the second Nationally Representative sample rated stories as significantly more difficult to tell than did participants in the first Nationally Representative sample, regardless of trauma type or ending.

Our moderation hypothesis was partially supported for perceived likelihood of sharing (see Table 4 and Figure 1b). Redemptive endings boosted the perceived likelihood of telling both IPV and non-IPV stories, but the effect of redemptive ending was stronger for IPV (vs. non-IPV) stories, opposite of our hypothesized direction of effect. Regarding covariates, those identifying as men rated stories as significantly less likely to be told than those of other self-identified genders. There was a significant main effect of participant lifetime trauma history; as participant self-reported trauma history increased, they rated the storyteller as more likely to share the story. Finally, participants in the two Nationally Representative samples rated the stories as significantly more likely to be told than participants in the University sample.

Our moderation hypothesis was not supported for storyteller likeability (see Table 5). There was no interaction between trauma type and story ending in predicting storyteller likeability. Instead, there was a main effect of story ending, such that storytellers who told stories with redemptive endings were rated as more likeable than those who told stories with negative endings, regardless of trauma type. No covariates were statistically significant predictors of storyteller likeability.

Our moderation hypothesis was partially supported for storyteller positive personality traits (see Table 5 and Figure 1c). Redemptive endings made audiences perceive both IPV and non-IPV storytellers as having more positive personality traits, but the effect of redemptive story endings was weaker for IPV stories. In other words, redemptive endings did not increase perceived positive personality traits of IPV storytellers as much as they increased perceived positive traits of non-IPV storytellers. There was a significant main effect of participant lifetime trauma history; as participants self-reported trauma history increased, they rated the storyteller as having less positive personality traits.

Our moderation hypothesis was not supported for empathy towards the storyteller (see Table 6). There was no interaction between trauma type and story ending in predicting empathy towards the storyteller, and there were no main effects of trauma type or story ending on empathy. Across all experimental conditions, the average strength of self-reported empathic emotions for the storyteller was about 5.35 on the 7-point scale from 1 (not at all) to 7 (very much). Consistent with the covariate effects in the tests of H1a, those identifying as men reported significantly less empathy for the storyteller than those identifying as other genders, and participants in the two Nationally Representative samples reported significantly more empathy for the storyteller than participants in the University sample.

Our moderation hypothesis was not supported for stigma towards the storyteller (see Table 6). There was no interaction between trauma type and story ending in predicting stigma. Consistent with the results of H1a, there was a main effect of trauma type, such that participants reported less stigmatizing attitudes toward IPV storytellers than non-IPV storytellers, regardless of story ending. Regarding covariates, there was a significant main effect of participant lifetime trauma history; as participant self-reported trauma history increased, they endorsed less stigma toward the storyteller. There was no effect of sample or male gender identity on stigma.

We hypothesized that audiences would perceive psychological (versus sexual or physical) violence stories with negative endings as more difficult to be shared and less likely to be shared; their storytellers as less likeable and having fewer positive personality traits; and participant responses would be less empathetic and higher in stigmatizing attitudes towards the storyteller. To test this hypothesis, the IPV story (with negative ending) variable was coded as 1 = psychological, 2 = sexual, 3 = physical, with a Helmert contrast comparing levels 1 vs. 2 and 3, and levels 2 vs. 3. Factorial ANOVA results for the test of this hypothesis in the combined sample can be found in Table 7.

Our hypothesis was supported for participant empathy for the storyteller (η²_p = .024). Specifically, participants endorsed significantly less empathy for the psychological violence storyteller with a negative ending, than the sexual or physical violence storyteller with a negative ending (M_D = -0.32, SE_D = 0.15, p = .037, 95% CI [-0.61, -0.02]). Our hypothesis was not supported for perceived difficulty sharing, likelihood of sharing, storyteller likeability, storyteller positive personality traits, or stigmatizing attitudes toward the storyteller.

The sample was a statistically significant predictor of perceived likelihood of sharing, where the Nationally Representative samples (versus the University sample) rated the IPV storytellers as significantly more likely to share their story (η²_p = .039), as was the case for the test of Hypothesis 1a. Consistent with the test of Hypothesis 1a, participants’ own self-reported history of traumatic life events was not a statistically significant predictor of any IPV storyteller ratings or IPV story perceptions, and male-identified participants endorsed significantly less empathy for the storyteller (η²_p = .028) than participants of other self-identified genders, regardless of IPV story type.

We left as exploratory whether there would be differences between how audiences perceived sexual violence stories with negative endings versus physical violence stories with negative endings. The only statistically significant difference in audience perceptions of sexual (versus physical) violence stories with negative endings was for perceived difficulty of sharing. Audiences perceived sexual violence stories with negative endings as significantly more difficult to share than physical violence stories with negative endings (M_D = 0.27, SE_D = 0.11, p = .020, 95% CI [0.04, 0.49]).

The pattern of results across our planned tests of how story perceptions would differ for psychological, sexual, and physical violence stories with negative endings suggests that the greatest differences in perception are between sexual versus psychological and physical violence stories—not, as we had hypothesized, between psychological versus sexual and physical violence stories. As such, we performed a data-driven, post-hoc exploratory analysis of H2a, this time with the IPV story (with negative ending) variable coded as 1 = sexual, 2 = psychological, 3 = physical, with a Helmert contrast comparing levels 1 vs. 2 and 3, and levels 2 vs. 3. All post-hoc exploratory analyses in the combined sample continued to control for sample, participant lifetime trauma history, and participant gender identity.

Post-hoc contrasts revealed that participants rated sexual violence stories with negative endings as significantly more difficult to share (η²_p = .039, M_D = 0.31, SE_D = 0.07, p < .001, 95% CI [0.17, 0.45]) and less likely to be shared (η²_p = .012, M_D = -0.24, SE_D = 0.10, p = .015, 95% CI [-0.44, -0.05]) than psychological and physical violence stories with negative endings. Additionally, participants reported more empathy for sexual violence storytellers, versus psychological and physical violence storytellers (M_D = 0.39, SE_D = 0.10, p < .001, 95% CI [0.19, 0.59]), and psychological violence storytellers as significantly less empathetic than physical violence storytellers (M_D = -0.22, SE_D = 0.11, p = .049, 95% CI [-0.44, -0.001]; η²_p = .036).

We hypothesized that narrative redemption would moderate the effect of IPV trauma type (sexual, physical, psychological) on story perceptions, such that redemptive story endings would enhance the positivity of audience perceptions of sexual and physical violence stories to a greater degree than psychological violence stories (we left as exploratory the contrast between sexual vs. physical violence stories). To test this hypothesis, the IPV trauma story variable was coded as 1 = psychological, 2 = sexual, 3 = physical, with a Helmert contrast comparing levels 1 vs. 2 and 3, and levels 2 vs. 3.

Contrary to our prediction, narrative ending did not moderate the effect of IPV trauma type (sexual, physical, psychological) on perceived difficulty sharing (see Table 8). Instead, there were main effects of IPV trauma type and story ending on perceived difficulty of sharing the story. Participants rated sexual violence stories as significantly more difficult to share than physical violence stories, regardless of story ending (this was a planned exploratory contrast). Participants also rated IPV stories with negative endings as significantly more difficult to share than IPV stories with redemptive endings, regardless of IPV trauma type. Finally, participants in Nationally Representative Sample 2 rated stories as significantly more difficult to share than did participants in Nationally Representative Sample 1.

Our moderation hypothesis was not supported for likelihood of sharing (see Table 8). Instead, there was a main effect of IPV story ending (redemptive vs. negative) on perceived likelihood of sharing the story. Participants rated IPV stories with redemptive endings as significantly more likely to be told than IPV stories with negative endings, regardless of IPV type. Moreover, participants in Nationally Representative samples 1 and 2 rated stories as significantly more likely to be told than did participants in the University sample.

Our moderation hypothesis was supported for storyteller likeability (see Table 9 and Figure 2a). Overall, audiences rated redemptive-ending (vs. negative-ending) IPV storytellers as more likeable, but this effect was particularly pronounced for sexual and physical violence storytellers. In other words, redemptive story endings enhanced the likeability of sexual and physical violence storytellers significantly more than psychological violence storytellers. No covariates, including sample, were statistically significant predictors of storyteller likeability.

There was weak support for our moderation hypothesis for storyteller positive personality traits (see Table 9). The moderator effect was marginally significant, suggesting that redemptive endings enhanced the likeability of sexual and physical violence storytellers more than for psychological violence storytellers. Averaged across IPV type, there was a main effect of story ending, such that participants rated IPV storytellers as having more positive personality traits when the stories ended redemptively (versus negatively). No covariates were statistically significant predictors of storyteller likeability.

Our moderation hypothesis was supported for empathy towards the storyteller (see Table 10 and Figure 2b). Narrative ending moderated the effect of IPV trauma type on empathy for the storyteller, such that redemptive endings boosted participant empathy for sexual and physical violence stories to a greater degree than for psychological violence storytellers. Visual inspection of the graph (confirmed by post-hoc follow-up analyses) suggests that this effect was driven by sexual violence storytellers with redemptive endings garnering particularly empathic responses from audiences. Regarding covariates, there were no effects of sample or participant lifetime trauma history on empathy for the storyteller, but those identifying as men reported significantly less empathy for storytellers, regardless of IPV type or story ending.

Our moderation hypothesis was not supported for stigmatizing attitudes towards the storyteller (see Table 10). Instead, there was a main effect of story ending, such that participant stigmatizing attitudes were stronger in response to stories with negative (vs. redemptive) endings, regardless of IPV type. Regarding covariates, there was a significant main effect of participant lifetime trauma history; as participant self-reported trauma history increased, their stigmatizing attitudes toward storytellers decreased, regardless of IPV type or story ending. Moreover, on average, participant stigmatizing attitudes were greater in the Nationally Representative samples than in the University sample.

We left as exploratory whether there would be differences between audiences’ perceptions of sexual versus physical violence stories. There were no statistically significant differences in audience perceptions of sexual versus physical violence stories, nor any statistically significant interaction of sexual vs. physical violence stories with story ending, with regard to perceived likelihood of sharing, storyteller likeability, storyteller positive personality traits, empathy, or stigmatizing attitudes towards the storyteller. However, as reported above, there was a statistically significant main effect of sexual versus physical violence story type in predicting difficulty of sharing.

The pattern of results across our planned tests of how IPV story perceptions would differ for psychological, sexual, and physical violence stories suggests there were differences not only—as hypothesized—between psychological versus sexual and physical violence stories, but also between sexual as compared to psychological and physical violence stories. As such, we performed a data-driven, post-hoc exploratory analysis of H2b, this time with the IPV story variable coded as 1 = sexual, 2 = psychological, 3 = physical, with a Helmert contrast comparing levels 1 vs. 2 and 3, and levels 2 vs. 3. All post-hoc exploratory analyses in the combined sample continued to control for sample, participant lifetime trauma history, and participant gender identity.

Post-hoc contrasts revealed that participants rated sexual violence stories as significantly more difficult to share than psychological or physical violence stories, though this effect did not depend upon ending (see Figure 3a). Additionally, there was a significant moderation effect of IPV type (psychological vs. physical) and story ending on ratings of storyteller likeability. The effect of story ending on storyteller likeability ratings was particularly pronounced for physical (vs. psychological) violence storytellers (see Figure 3b). In other words, having a negative ending made physical violence storytellers particularly disliked (compared to psychological violence storytellers), while having a redemptive ending made physical violence storytellers particularly liked (compared to psychological violence storytellers). The visual pattern of results for storyteller positive personality traits was similar to the likeability findings, but the moderation effect was only marginally significant. However, there were main effects of both story ending (redemptive vs. negative) and IPV type (psychological vs. physical), such that psychological violence storyteller personalities were rated more positively than physical violence storyteller personalities. For empathy toward the IPV storyteller, there were significant main effects of both story ending (redemptive vs. negative) and IPV type (sexual vs. psychological and physical). Participants reported significantly more empathy for the sexual (vs. psychological and physical) violence storyteller, regardless of story ending.

There were no differences between sexual (vs. psychological and physical) or between psychological versus physical violence stories in perceived likelihood of sharing or stigmatizing attitudes toward the storyteller.

For ease of reference, Tables 11 and 12 present at-a-glance summaries of all study findings.

The adult participants in this vignette-based experimental study rated stories of interpersonal violence (IPV), versus non-IPV, as more difficult to share and less likely to be shared, and their storytellers as having less positive personality traits. Effect sizes were medium-to-large. This core finding on perceptions of IPV versus non-IPV stories replicates and extends the generalizability of research by McLean et al. (2020) and Delker et al. (2020), who showed more stigmatizing audience reactions to stories of sexual violence (adult sexual assault and childhood sexual abuse) versus impersonal, accidental traumas. Further replicating the prior work, audiences consistently rated trauma stories and their storytellers more favorably when the stories ended with personal redemption. Contrary to our hypotheses about IPV story perceptions, psychological violence did not emerge as a consistently or robustly more stigmatized form of IPV than sexual or physical violence. Instead, audiences expected that sexual violence stories would be particularly difficult and unlikely to share. Below we review the results of our study in more detail, and we consider practical implications for trauma storytellers, along with methodological suggestions for future research on cultural stigma surrounding IPV.

Redemption is a culturally valued way to conclude a story of adversity (e.g., McAdams, 2006; McLean et al., 2020), but we had anticipated that redemptive endings would enhance favorable responses only to non-IPV stories, given the pervasive cultural stigma surrounding IPV. Our results provided partial support for this prediction: redemptive endings boosted the perceived ease of telling all trauma stories and the perceived positive personality traits of all trauma storytellers, but the boosting effect was stronger for non-IPV (versus IPV) storytellers. These findings conceptually replicate the findings of Delker and colleagues (2020) with—in the present study—longer, more detailed trauma stories, all focused on adulthood, and with the three major types of IPV rather than sexual violence only. Despite the culturally valued story structure of a positive ending to adversity characterized by personal resilience, the stigma of experiencing harm within a trusted ongoing relationship (as opposed to experiencing impersonal traumas) seems to stick to IPV storytellers (Adams-Clark et al., 2020; Delker, 2021). We use the term ‘stigma’ with intention here, in that the more negative, culturally devalued personality traits attributed to the IPV (versus non-IPV) storytellers represent a mark on or devaluation of the person herself, in consequence of having certain experiences. Contrary to our hypothesis in this study and to prior findings by Delker et al. (2020), redemptive endings boosted the perceived likelihood of sharing IPV stories more than they boosted the perceived likelihood of sharing non-IPV stories. Given that these findings contrast with those of our prior study with similarly representative samples (Delker et al., 2020), we are reluctant to interpret this without further decisive testing. A study design that asks open-ended questions of participants, such as a qualitative interview-based study or focus group design, could start to unpack the more nuanced reasons why audiences perceive stories with redemptive endings to be more or less likely to be shared.

Several additional predictions about IPV versus non-IPV storytellers were unsupported. First, participants did not differ in their empathy or liking for IPV versus non-IPV storytellers, regardless of ending. An explanation of the unexpected null finding for empathy is that participant and sample characteristics outweighed our experimental manipulation of trauma type in explaining empathic reactions to storytellers. In our mega-analytic tests of hypotheses 1a and 1b, the only statistically significant predictors of empathy were participant gender (males reported significantly less empathy for the storytellers, a small-to-medium effect size) and study sample (the two nationally representative samples reported significantly more empathy for storytellers than did the university student sample, a small effect size). The gender finding is consistent with survey-based research on dispositional empathy that has found less self-reported expression of empathic concern among more masculine identities, though far more research is needed to understand the nature, extent, and potential malleability of men’s empathy for trauma survivors (Burris et al., 2016; Osman, 2011). The finding of higher empathy among the two nationally representative samples may be explained by the softening effect of their broader life experience (average age in middle adulthood) compared to the university sample (average age in emerging adulthood) or to the prevalence of rape myth acceptance on college campuses (Holland et al., 2020; Rich et al., 2021). Future research on trauma story perceptions might be able to detect effects of trauma stories on empathy with stronger, more visceral experimental manipulation of trauma stories, or more complex, multi-modal measurement of empathic concern (Dawtry et al., 2020).

Support was mixed for our hypothesis predicting more negative perceptions of psychological (versus sexual and physical) violence storytellers. In partial support of the hypothesis, redemptive endings boosted participant liking of and empathy toward all IPV storytellers, but the boosting effect was more pronounced for sexual and physical (versus psychological) violence storytellers. In addition, compared to sexual and physical violence storytellers, psychological violence storytellers garnered the least empathy from participants overall, regardless of story ending. However, these empathy findings are qualified by the relatively strong effect of participant/sample characteristics on empathy (summarized in the previous section), and by ceiling effects in the measurement of empathy: across all trauma types, participant responses averaged above a 5 on a 7-point scale for feeling empathic emotions (from 1, not at all to 7, very much).

Our hypothesis predicting more negative perceptions of psychological (versus sexual and physical) violence storytellers was not supported for perceived difficulty sharing, perceived likelihood of sharing, storyteller positive personality traits, nor stigma toward the storyteller. Instead, post-hoc exploratory analyses showed that sexual (versus psychological and physical) violence stories were perceived as less likely to be shared and more difficult to share, with small-to-medium effect sizes. Post-hoc analyses also showed that the trend for less participant empathy toward psychological violence storytellers was driven by the particularly high degree of empathy for sexual (versus psychological and physical) violence storytellers, a medium effect size. We had anticipated that the more diffuse, invisible nature of psychological violence—which tends not to leave overt, concrete evidence of harm—would spur particularly negative audience reactions to these storytellers. Instead, it is possible that the very features of psychological violence that enable bystanders to minimize its harmfulness in real time may explain dampened participant reactions to psychological violence storytellers in this study, such as lower participant empathy and null effects for perceived difficulty of sharing the psychological violence story.¹

Pronounced participant reactions to sexual (versus psychological and physical) violence storytellers in this study might be explained by the outsized influence of the #MeToo movement on mainstream discourse about IPV. As a result of absorbing approximately five years of #MeToo stories shared across social media and other public platforms, the adults in this study may be more sensitized to the difficulties of telling sexual violence stories and to the many understandable barriers to disclosing or formally reporting an experience of sexual violence (Lorenz et al., 2019; Onwuachi-Willig, 2018; Stubbs-Richardson et al., 2018). If this were the case, participants may have drawn on an availability heuristic of sorts in their recognition of how hard it is to share a sexual violence story, relative to psychological and physical violence stories, which have been largely absent from public discourse about IPV. Whereas we had anticipated that features of the IPV itself (i.e., the overt physical harm of choking and forced vaginal penetration, versus the covert harm of psychological violence) would spur differential audience reactions across IPV types, our findings allude to the possibility that mainstream, dominant cultural representations of IPV trauma in the U.S. have the greatest influence on audience reactions to IPV storytellers.

The overall trend across our four hypotheses was a null effect of trauma stories on participant ratings of storyteller likability and stigmatizing attitudes toward the storyteller. The only modest exceptions were the small moderation effect of redemptive endings on storyteller likability for the IPV stories, discussed above, and a small effect of IPV (versus non-IPV) trauma type on lower stigmatizing attitudes, contrary to the hypothesized direction of effects. The null effect of trauma stories on perceived storyteller likability deviates from the results of McLean et al. (2020) and Delker et al. (2020), but so too did the internal consistency reliability of the storyteller likability items in the present study, which was adequate but surprisingly low compared to the aforementioned studies.

One intriguing explanation for the above findings on likability and stigma is a possible cohort effect in this study driven by the global COVID-19 pandemic. Over the past two years, the contagiousness of the virus, social distancing and quarantine requirements, and work and school closures have had an enormous impact on social life and perceptions of risk in everyday interactions. Yet the published likability and stigma measures used in this study, which were designed prior to the COVID-19 pandemic, tap into respondent willingness to interact with the storyteller in social settings and at work (e.g., “work closely on a job with the author of this story”). Indeed, the published stigma self-report measure used in this study operationalizes stigma as desired social distancing from the protagonist. Given that two of the three non-IPV vignettes mention hospital stays for the sick/injured storyteller, it is perhaps not surprising that participants in this study reported stronger social distancing (stigma) responses toward the non-IPV storytellers, than the IPV storytellers.

Despite the strengths of our well-tested trauma story vignettes and varied samples of U.S. adults, we consider several methodological and practical limitations to this work. First, we measured the story perception variables (perceived difficulty and likelihood of sharing) with one-item measures. There is a precedent for using one-item measures in large-scale survey-based research on face-valid constructs (e.g., Cheung & Lucas, 2014), but future research can explore more varied dimensions of story perceptions with expanded or multi-method measures. Second, as explained above, measurement challenges in accurately assessing storyteller likability and stigmatizing attitudes in this COVID-19-era study point to the need for additional research before firm conclusions can be drawn about how trauma type influences these reactions.

The third issue represents an inherent practical challenge to highly controlled experimental research on trauma storytelling (Syed, 2021). We carefully designed the experimental manipulation of IPV trauma type in this study, to make each of the three IPV stories identical (except for the description of the traumatic events themselves). But psychological violence is an inherently more diffuse form of IPV, one whose emotional and existential harm perhaps cannot be captured adequately in as few words as can the acute harm and horror of a sexual or physical assault. As such, by prioritizing internal validity in our experimental design process and presenting equivalently structured stories of psychological, sexual, and physical violence, we may have asked audiences to compare apples to oranges. This conundrum for experimentalists points to the need for descriptive and observational research on psychological violence storytelling that explores survivor experiences in trying to put words to a more diffuse and less culturally recognized form of harm. Follow-up studies can employ “structural-psychological” research methods that are sensitive to how individual self-understanding and storytelling processes are intertwined with dominant cultural values and power structures (Syed & McLean, 2021, p. 1). Event narratives, conversation analysis, focus groups, in-depth interviews, life-script analysis, and participatory methods such as digital storytelling are among the promising mixed-method and qualitative forms of inquiry into IPV storytelling experiences (Fish & Syed, 2021; Syed & McLean, 2021, p. 1) 

Before empathy toward psychological violence survivors can be evoked and momentum for systemic change sustained, perhaps broader awareness and understanding are needed—and a depiction the length of several Tweets will not accomplish this understanding. Broader understanding of IPV is important for many reasons, but we follow Salter and Hall (2020) to highlight how greater understanding and cultural validation could serve as a dignifying antidote to the shame and humiliation of gender-based violence. Trauma healing/recovery is often framed as an individual psychological journey for survivors, but to heal from profound shame and injury (and to prevent long-term, complex posttraumatic stress conditions), survivors would benefit from environments that are better sensitized to the inherent “value and vulnerability of all living things” (Salter & Hall, 2020, p. 7)—and better equipped to recognize the dynamics of IPV.

Overall, this study is part of growing awareness and empirical attention paid toward the challenges of storying particular types of events, and toward the role that social perceptions, cultural stigma, and dominant cultural expectations play in the storytelling process. Contrary to mainstream American ideology that uplifts personal agency and autonomy in self-authorship, our approach highlights structural constraints to self-authorship. We hope that this approach will inspire more work to tease apart how these multiple factors work together to silence or validate particular stories and particular people.

B.C.D. and K.C.M. conceptualized the study, designed the first draft of the study methodology, provided project administration and supervision for the study, and acquired institutional financial support for this project. All authors (B.C.D., K.C.M., P.K.M., K.T.) contributed to the final version of the study design and hypotheses. B.C.D. acquired the pilot data (Stage 1). K.T. acquired the 3 samples of study data for the registered report study (Stage 2). B.C.D., P.K.M., and K.T. processed the raw data, curated the raw data, created the data analysis syntax, and performed the formal data analysis for both the pilot study and the registered report study. B.C.D., K.C.M., P.K.M., and K.T. contributed to the interpretation of the data presented in this Stage 2 Registered Report manuscript. B.C.D. wrote the initial drafts of the manuscript in Stages 1 and 2, and all co-authors provided critical commentary and revisions. All authors approved the final version of the paper for submission.

Hannah Wolin contributed to revising the study experimental materials and to online pilot data collection in Stage 1. Thea Huber, Annalyn Ledesma, and Greyson Mize contributed to creating tables summarizing the mega-analytic results.

This registered report study was funded by a Collaborative Research Project award from Western Washington University’s Center for Cross-Cultural Research (#CRC_IPV) awarded to principal investigators B.C.D. and K.C.M.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Following Stage 1 in principle acceptance, we registered our approved protocol on the Open Science Framework under private embargo until submission of the Stage 2 manuscript. Our approved Stage 1 protocol, anonymized raw data, digital study materials (including stimuli, experiment code, and analysis code) and laboratory log for all published results are publicly accessible on OSF at the following link: https://osf.io/z76ue/