Abstract
Intimate partner violence (IPV) is a prolific and growing issue that can have long-lasting negative consequences for the health and safety of those involved. Bystander intervention is one method for helping to combat incidents of IPV, as research suggests that bystanders are frequently present at the scene of assaults and incidents of IPV. This study explored individual differences of bystanders that may influence whether they decide to intervene in an unfolding incident of IPV, as well as how the likelihood of intervening may vary as a function of the apparent gender or sexual orientation of the individuals involved in an incident of IPV. Participants were recruited from an online survey platform to obtain a balanced sample of heterosexual and sexual minority individuals. Participants completed a bystander task where they listened to an audio vignette of an unfolding IPV incident and were instructed to stop the audio if/when they would intervene in a real-life context. Participants were randomly assigned to listen to one of four versions of the vignette in which the apparent gender of the aggressor and victim were manipulated. Results revealed participants were more likely to intervene if they identified as a sexual minority (vs. as a heterosexual), reported less rape myth acceptance, or had greater endorsement of gender equality. Results also revealed that associations between bystander characteristics and intervening behavior largely did not differ across vignette conditions, suggesting that they may influence the likelihood of intervening consistently across incidents of IPV regardless of the apparent gender and sexual orientation of the aggressor and victim. However, participants in general were most likely to intervene in the male aggressor/female victim vignette. Implications for IPV prevention programming—including the need to include more diverse and less heteronormative depictions of IPV—are discussed.
Intimate partner violence (IPV) is a prolific and growing issue that poses significant health and safety risks for many individuals in the United States (National Coalition of Anti-Violence Programs, 2016). IPV refers to any physical, sexual, or psychological violence, or stalking behavior by an intimate partner. These behaviors can span anywhere from physical aggression and verbal abuse, to controlling a partner’s money or access to support systems. More than one in three women and one in four men in the United States have experienced rape, physical violence, or stalking at the hands of an intimate partner during their lifetime (Smith et al., 2018). IPV is associated with numerous long-term consequences for the survivor, such as increased rates of depression and anxiety, poor physical health, and difficulties with future interpersonal relationships, amongst many others (Capaldi et al., 2012; Smith et al., 2018).
With incidents of IPV continuing to increase at alarming rates (Rennison & Welchans, 2000; Smith et al., 2018), it is imperative to identify and understand the possibilities for prevention of escalating violence. Indeed, the increase in incidents of IPV in the United States has been especially pronounced over the past few years during the COVID-19 pandemic (Kofman & Garfin, 2020; Piquero et al., 2021), with increases in IPV speculatively linked to changes in financial insecurity, the stress of childcare, and maladaptive coping strategies throughout the pandemic (Piquero et al., 2021). The circumstances surrounding the pandemic and its mitigation strategies, such as isolating at home, also likely resulted in reduced access to potential sources of support or aid outside the home and reduced the likelihood that a bystander would be present during incidents of IPV.
In the present study, we aimed to examine the role of bystanders in preventing incidents of IPV, with a focus on how individual differences of the bystander may influence whether or not they decide to intervene, including their sexual orientation, previous victimization history, and general beliefs about gender roles and sexual assault. Critically, we examined bystander responses to unfolding fictitious incidents of IPV where the perceived sexual orientation and gender of the parties involved in the incident were systematically manipulated, allowing us to contribute to an area of research where there is limited empirical data at present.
Bystander Characteristics and Beliefs
Previous sexual violence prevention programming focused on risk-reduction strategies, targeting facets of both victim and perpetrator behavior (McMahon, 2010). Examples include behavior modification, self-defense, and assertiveness training (Franklin et al., 2017). Although these methods are still employed and have demonstrated some success, other forms of prevention programming have shifted toward a community-oriented model, engaging community members as resources for both prevention and recovery (McMahon, 2010). Bystanders, who are often present to witness acts of IPV (Franklin et al., 2017), have the opportunity to intervene and prevent further victimization and harm. The call for bystander intervention has placed a greater responsibility on individual bystanders and members of the community as a whole to play a role in primary prevention, interrupt acts of violence when witnessed, and help support survivors following an experience of victimization.
A bystander is a witness to a crime, emergency, or high-risk situation, who is not themselves directly involved as a perpetrator or a victim (Banyard & Moynihan, 2011). Data from the National Crime Victimization Survey has shown that bystanders are present at over 70% of all physical assaults and around 29% of reported rapes, making them an incredibly important asset in preventing high-risk interpersonal situations (Franklin et al., 2017). McMahon (2010) suggests that this holds especially true in “at-risk” environments like college settings, where most assaults are committed by someone known to the victim and occur in social settings such as residence halls or fraternities with others present. It has been demonstrated that in these environments, bystanders are typically present during the “pre-assault phase” where red flags begin to appear and, if equipped with the appropriate skills, can intervene to disrupt the escalation of danger (McMahon, 2010). As a result, bystander intervention can be an incredibly powerful tool in preventing the occurrence of IPV when implemented. In addition, the increasing utilization and research of bystander intervention programs such as “Green Dot” and “Bringing In The Bystander” have found this method of reducing sexual assault to be effective in a variety of settings such as in high schools and on college campuses (Coker et al., 2011; Edwards et al., 2019; Mujal et al., 2021). However, there are several factors that may potentially influence a bystander’s decision to intervene during or after witnessing an incident of IPV.
Endorsement of gender norms
Gender norms are social norms that relate specifically to gender differences, defining or prescribing acceptable and appropriate behaviors for members of a society, and can be found in both formal and informal institutions (Overseas Development Institute, 2015). Gender norms stem from gender stereotypes, such as the characterization of men as strong and tough, and women as emotional and domestic (Davies, 2002; Sorenson & Thomas, 2009). Personal acceptance or perceived acceptance of gender norms can negatively influence bystander behavior, as the bystander may not view the victim as needing or deserving help (Brown & Messman-Moore, 2010; Salter et al., 2021). For example, bystanders are generally more likely to intervene for victims who are stereotyped as being more feminine or “weaker” and needing more help (Sorenson & Thomas, 2009). Acceptance of gender norms can also lead to the belief that only women can be victims and men can be perpetrators and, as a result, create the mindset that only violence against women is serious enough to take action (Seelau et al., 2003). Contrastingly, the stereotype of women as dainty and ladylike may lead one to believe that if a woman was victimized, she must have done something wrong to provoke an attack, such as breaking traditional prescribed gender norms (Seelau et al., 2003). For example, those who maintain sexist gender role stereotypes are more likely to believe it is natural for men to dominate women and tend to hold more negative attitudes toward female rape victims (Davies et al., 2012).
In the case of men being raped, those with sexist attitudes and an endorsement of gender norms were more likely to endorse beliefs that the man must have acted in an “unmanly” fashion, which provoked an attack (Davies et al., 2012). These attitudes arise from the traditional view of masculinity where men are seen as strong, assertive, sexually dominant, and heterosexual, and suggest that men “cannot” be raped. As a result of these beliefs, the rape of a man may not be taken as seriously as the rape of a woman among those who more strongly endorse traditional gender norms. Or, it may be assumed that even though a sexual interaction was not consensual, it should be viewed positively by the male victim based on the false belief that men are always looking for sexual experiences. Such beliefs can lead to a lesser likelihood of intervention (Davies, 2002).
Rape myth acceptance
The term “rape myth” refers to “attitudes and beliefs that are generally false yet widely and persistently held and that serve to deny and justify male sexual aggression against women” (McMahon, 2010, p. 4). Some examples of rape myths include, “when women go around braless or wearing short skirts and tight tops, they are just asking for trouble” and “many women secretly desire to be raped” (Edwards et al., 2011, p. 762). Although the content of rape myths are untrue, they continue to exist in a large segment of the population and legal, media, academic, and religious institutions (Overseas Development Institute, 2015); research suggests that between 25% and 35% of respondents, both male and female, positively endorse the majority rape myths, though men are more likely to endorse rape myths than are women (Edwards et al., 2011; Lonsway & Fitzgerald, 1994).
Critically, a study conducted by McMahon (2010) found that self-reported rape myth acceptance was a negative predictor of self-reported bystander attitudes. If a bystander believes that someone is responsible for their own assault, they will be less likely to intervene as they see the victim as deserving or even welcoming of the attack (Pagliaro et al., 2021). It has also been reported that greater endorsement of rape myths is associated with a greater number of perpetration behaviors related to sexual violence (Banyard & Moynihan, 2011). In addition, those with a higher acceptance of rape myths may have lower rates of intervention as they do not consider the witnessed behavior to be aggressive or dangerous (Leone et al., 2021).
Prior unwanted sexual experiences
Research has found that an increased sense of empathy and efficacy may result in more positive bystander behavior and greater likelihood of intervention (Leone et al., 2021; McMahon et al., 2020). When a situation is less ambiguous or a bystander has a greater knowledge of IPV warning signs to compensate for ambiguity, bystander intervention is more likely to take place. In a study examining historically underrepresented students, it was found that experiencing forms of oppression, discrimination, and abuse was associated with an increase students’ willingness to intervene in similar situations (McMahon et al., 2020). Consequently, it is likely that those who have a past experience with unwanted sexual activity of some kind (e.g. previous experience of sexual assault or IPV) will be more apt to notice incidents of sexual and dating violence and will be more willing to step in to help, as the perceived benefits of doing so may outweigh the perceived consequences. For example, Chabot et al. (2009) found that participants with a history of childhood abuse were more likely than not to intervene in a domestic violence incident and suggested that for young adults with limited experience “the overall amount of abuse experiences, not the type, is more influential” in decisions to intervene (Chabot et al., 2009, p.1707).
Gender and sexual orientation
Extensive research has been conducted on gender differences in bystander behavior, and women are generally more likely to intervene as bystanders and show more concern for the victim than do men (Bennett et al., 2017; Seelau et al., 2003). Gender differences in bystander behavior may be linked to increased recognition or efficacy of warning signs, increased empathy for the victim, and/or lack of endorsement of traditional gender norms and rape myths (Davies, 2002; National Coalition of Anti-Violence Programs, 2016; Salter et al., 2021). These potential explanations of gender differences in bystander behavior may also apply to other individuals from sexual and gender minority groups (Hoxmeier et al., 2022). For example, women experience IPV at rates higher than that of men (Bennett et al., 2017; Rennison & Welchans, 2000), as do individuals who identify as sexual and/or gender minorities (Edwards et al., 2015; National Coalition of Anti-Violence Programs, 2016), and thus individuals from these groups are subsequently more likely to have or know someone who has had a prior victimization than are cisgendered, heterosexual men.
Perceptions of IPV in LGBQ+ Couples
The existence of homophobia and sexism within the United States makes it difficult for lesbian, gay, bisexual, queer, or other (LGBQ+) identifying individuals to report IPV and ask for help, but also leads to less intervening behavior in incidents of IPV involving non-heterosexual individuals overall (Bornstein et al., 2006; Jin, 2017; Pagliaro et al., 2021). In reported cases of IPV, gay male victims are blamed more for their abuse and it is considered less serious than for other victims, especially for bystanders with greater homophobic beliefs and greater rape myth acceptance (e.g. that homosexuality is unnatural and gay men deserve rape; Davies et al., 2012). Moreover, reactions toward a victim, regardless of gender, depend on their sexual orientation; victims are blamed more when the perpetrator is a member of the gender that the victim is normally attracted to. For example, male heterosexual victims were blamed more when the perpetrator was female than when male, whereas gay male victims were blamed more when the perpetrator was male than female (Davies, 2002).
For some bystanders, likelihood of intervening in an incident of IPV in a LGBQ+ relationship may also be lower due to a lack of understanding of non-heterosexual relationships and the manifestation of abuse within them (Davies, 2002). Studies have demonstrated that police were less likely to arrest a perpetrator in a case not involving male-against-female violence (Pagliaro et al., 2021; Pattavina et al., 2007). This reiterates the point that for individuals who are unaware of the dynamics and intricacies of a same-sex relationship, it may be easy to overlook the warning signs of IPV. In lesbian relationships, individuals with homophobic beliefs are likely to view the partnership as just temporary experimentation or as only a strong friendship, not a romantic relationship (Davies et al., 2012). This perception reduces the potential seriousness of the matter and makes it more likely that a bystander will not intervene when witnessing aggressive or abusive behavior.
Present Study
In this study, we were interested in examining bystander intervention behavior during an unfolding (fictitious) incident of IPV, with a focus on individual differences associated with increased likelihood of a bystander intervening across IPV incidents that varied in terms of the apparent gender and sexual orientation of the victims and aggressors. Participants, who were purposefully sampled to be roughly half heterosexual identifying and half LGBQ+ identifying, were presented with a realistic audio vignette depicting an unfolding incident of IPV where they were provided with an opportunity to intervene. Participants were randomly assigned to one of four conditions in which the apparent gender of the victim and aggressor in the vignette were manipulated (male aggressor, female victim; male aggressor, male victim; female aggressor, female victim; and female aggressor, male victim). We predicted that individuals would be more likely to intervene if they reported less rape myth acceptance, reported less gender norm endorsement, reported a prior experience of intimate partner or sexual violence, were women (compared to men), and were LGBQ+ (compared to heterosexual). Additionally, we predicted that participants would, in general, be more likely to intervene in the condition where the IPV incident involved an apparent male aggressor and female victim compared to in all other conditions.
To date, most studies of this type have used self-report surveys of bystander attitudes or written vignettes describing an incident of IPV to assess self-reported bystander intervention intentions (Franklin et al., 2017; Palmer et al., 2018; Sorenson & Thomas, 2009). Written vignettes may not be as realistic as vignettes presented in other mediums, such as audio or video, which may limit their applicability for understanding real-world intervention behavior; for example, there is extensive evidence that people’s self-reported intention to intervene (or not) in response to a hypothetical written scenario may not be a particularly good indicator of actual intervention behavior (Marx & Gross, 1995; Witte & Kendra, 2010). By utilizing an audio vignette for this study and asking participants to actually “intervene” by stopping the recording, we hoped to increase the ecological validity of our bystander intervention task and thus its generalizability to real-world bystander behavior.
Method
This study was approved by the University of New Hampshire’s Institutional Review Board. All measures, manipulations, and exclusions in the study are disclosed. Data collection was closed prior to all analyses. The data and materials are available free and open to the public on the Open Science Framework (https://osf.io/h2dfb/).
Sample Size
An a priori power analysis conducted in G*Power (3.1.9.2) revealed that a sample size of N = 84 would provide sufficient power (80%) for detecting a medium effect size (r = .30) for investigating associations between intervening behavior and individual difference variables within any of the four audio vignette conditions separately (total N = 336). To account for potential data loss that can occur in online studies (e.g. missing comprehension or attention checks), we aimed to collect a total of 400 participants.
Participants
Four hundred forty-seven participants were recruited from Prolific (Prolific.co), an online participant recruitment platform, to participate in the study. Inclusion criteria for the study included being between the ages of 18 and 45 years old, fluent in English, currently living in the United States, and having access to a computer with audio output with headphones. Separate advertisements for the study were posted on Prolific: one restricted only to LGBQ+ and one restricted to only heterosexual individuals. Participation was via self-selection. Prior to analyses, 45 participants were excluded because they failed to complete the study, and an additional 14 participants (14/402, or 3%) were excluded because they failed both comprehension check questions (see below). This left a final analytic sample of 388 participants ranging in age from 18 to 45 (Mage = 27.3, SDage = 7.0). Of this final analytic sample, 43.3% identified as a man or as male, 51.3% identified as a woman or as female, and 5.4% identified as either a trans man, trans woman, non-binary, two-spirit, questioning, or agender. Of the final sample, 51.5% identified as heterosexual. The sample was 67.8% White, 8.2% Black, 8.0% East Asian, 7.7% multiracial, and less than 5% each of South Asian, Native American, Pacific Islander, and other race(s) not listed; 12.1% of the sample identified as Hispanic or Latinx. Demographic characteristics for the sample as a whole as well as for each vignette condition separately are provided in Table 1. Participants were compensated $4 for completing the experiment, which they completed online in a single experimental session lasting approximately 20 to 25 minutes.
Demographic Characteristics: for Full Sample and by Condition.
Note. agg = aggressor; vic = victim.
“Another gender” includes individuals who identify as a trans man, trans woman, non-binary, two-spirit, questioning, or agender.
Totals do not sum to 100% because 10 individuals reported agender as their sexual orientation and were omitted from analyses on sexual orientation, 7 individuals preferred not to report their race, and 15 individuals preferred not to report their ethnicity.
“Another race” includes individuals who identify as South Asian, Native American, Pacific Islander, or other race(s) not listed.
Materials
Intimate partner violence vignette
A 4-minute audio vignette depicting an incident of IPV was presented to participants in which two dating-partners are engaged in conversation and some sexual contact, which progresses to nonconsensual touching and eventually to rape. The script was closely adapted from a vignette originally developed and reported on in prior research (Marx & Gross, 1995). There were four recordings of the same script, and participants were randomly assigned to listen to one:one with a male aggressor and a female victim, one with a male aggressor and a male victim, one with a female aggressor and a female victim, and one with a female aggressor and a male victim. The aggressor (Jordan) and the victim (Alex) were given gender neutral names and these names were consistent across all vignettes. All recordings used an identical script, and actors were instructed to try to perform the script in the same way across recordings to reduce variation as much as possible between the conditions. Across conditions, four actors portrayed all parts, with the same actor portraying the same role in the different conditions (e.g. a single female actor portrayed the female victim in both vignettes with a female victim, and a single male actor portrayed the male victim in both vignettes with a male victim). The full script of the vignette is available in Supplemental Appendix A, and all audio vignette stimuli are publicly available for download online (https://osf.io/h2dfb/).
Tasks and Measures
Bystander task
Participants first received instructions for completing the bystander task. They were instructed to imagine that they were listening to an uncomfortable interaction from the next room over and were given instructions on how to track their comfort level while listening to the interaction. Participants tracked how uncomfortable they thought the situation was by clicking anywhere along a yellow-to-red color gradient, where yellow represented feeling more comfortable and red represented feeling less comfortable. They were told to click on the gradient as many times as desired, anytime they felt their comfort level change, while they listened to the vignette. They were instructed to press a button to stop the recording if they reached a point at any time when they would intervene in the situation if it were occurring in real life. Participants were told that they would be given an opportunity at the end of the study to listen to the full audio vignette clip if desired. This was done because in prior work there was concern people would not intervene because they wanted to hear the rest of the story. Complete participant instructions are available in Supplemental Appendix B or online (https://osf.io/h2dfb/). Our primary outcome measure for the bystander task was whether each participant decided to intervene or not. We also recorded how long (in seconds) it took participants to intervene. The comfort level tracking data is not analyzed or reported in the present article but is freely available online for interested researchers (https://osf.io/h2dfb/).
After stopping the clip (or at the end of the clip for those who did not elect to intervene), participants were asked two multiple choice comprehension questions to ensure that they listened to and could hear the audio clip: (1) What were the two dating partners’ names? (2) What did the dating partners talk about in the very beginning of the clip? Responses to these items were used to ensure that participants were paying attention to the audio vignette. Participants who answered both questions incorrectly (n = 14; 3%) were excluded from the sample prior to analyses. We elected only to exclude individuals who missed both attention check questions as, particularly in the absence of explicit instructions to remember details from the vignette, we felt it possible for a listener to potentially miss one of these two items. However, we felt it would be very unlikely that anyone who had listened to the vignette would miss both.
Later, participants were asked four open-ended questions: (1) What would you do or say in the event you chose to intervene? (2) Why did you choose to intervene when you did? (3) Do you think you saw any red flags that tempted you to intervene sooner? If so, what were they (e.g. specific exchanges or words)? (4) Were there any perceived barriers to intervening? (e.g. was there a time when you thought you should intervene but did not yet stop the recording for any reason?). Responses to these open-ended questions are outside the scope of the present investigation, but all data are freely available online for interested researchers (https://osf.io/h2dfb/).
Perceptions of the vignettes
Participants were also asked six questions concerning their perceptions of the vignette they had listened to as a means of collecting normative evaluations of the vignettes, to support interpretation of results and the use of these audio vignettes in future studies. First, participants were asked the extent to which they agreed the overheard interaction would be categorized as sexual assault, with responses on a 7-point Likert scale anchored at 1 = “Strongly disagree” and 7 = “Strongly agree.” Later, they were asked to report the perceived gender of both the aggressor and victim in the vignette they heard, with response options of “woman,” “man,” and “unsure.” Then they rated how realistic they perceived the overheard interaction to be on a 7-point Likert scale anchored at 1 = “Very unrealistic” and 7 = “Very realistic.” Finally, they answered two questions of the form “How likely is it that someone in a similar situation to Jordan (aggressor) [Alex (victim)] would respond the same way as Jordan [Alex] did here?” Responses to these questions were also on a 7-point Likert scale anchored at 1 =“Very unlikely” and 7 = “Very likely.”
Questionnaires
Following completion of the bystander task, participants were asked to complete a set of questionnaires online to assess demographic information (e.g. age, race) as well as individual difference variables we predicted may be associated with bystander behavior.
Unwanted sexual activity
The Unwanted Sexual Activity Scale (Muehlenhard & Cook, 1988) assesses an individual’s sexual experiences including those which were nonconsensual. We included 24 items from this scale. Reliability for this scale and its subscales was not reported in the original article (Muehlenhard & Cook, 1988), but item factor loadings for the utilized subscales reported in the original article indicated acceptable reliability: physical coercion (range = 0.42–0.87), intoxication (range = 0.72–0.79), altruism (range = 0.41–0.74), termination of relationship (range = 0.35–0.77), partner’s verbal coercion (range = 0.37–0.70), sex-role concerns (range = 0.46–0.67), and partner’s threat of self-harm (0.71) (Muehlenhard & Cook, 1988, pp. 61–62). Convergent validity was demonstrated with significant associations between self-reported unwanted sexual activity and scales assessing attitudes toward women, sex-role stereotyping, adversarial sexual beliefs, and acceptance of interpersonal violence (Muehlenhard & Cook, 1988). Questions ask “Have you ever engaged in sexual activities (ranging from kissing to sexual intercourse) when you didn’t want to because. . .” with response options of 0 = “you never felt this way or encountered this situation at all,” 1 = “you felt this way or encountered this situation but never engaged in any unwanted sexual activity because of it,” 2 = “you went as far as kissing for this reason,” 3 = “you went as far as petting for this reason,” and 4 = “you went as far as sexual intercourse for this reason” (Muehlenhard & Cook, 1988, pp. 60–62). Response values of 2 to 3 were coded as indicative of having experienced unwanted sexual contact and response values of 4 were coded as indicative of having experienced unwanted sexual intercourse. Using this criteria, 195 participants (50.3%) reported experiencing at least one instance of unwanted sexual contact, 185 participants (47.7%) reported experiencing at least one instance of unwanted sexual intercourse, and 256 participants (66.0%) reported experiencing some form of unwanted sexual experience (either unwanted sexual contact or unwanted sexual intercourse).
Rape myth acceptance
The Rape Myth Index (Stoll et al., 2017) assesses individual’s endorsements of rape myths, including about individuals who are sexual or gender minorities, racially diverse, low socioeconomic status (SES), and/or sex workers. This measure, consisting of 28 items, was used to assess individual’s self-reported rape myth acceptance. Prior work has provided evidence of the scale’s validity (Stoll et al., 2017) and reliability (Cronbach’s α = .92; Stoll et al., 2017), and it also demonstrated good reliability in the present study (Cronbach’s α = .96). An example item is, “women tend to exaggerate how much rape affects them” (Stoll et al., 2017, p. 34). Participants were asked to rate their agreement with each of the 28 statements using a 5-point Likert scale ranging from 1 = “strongly disagree” to 5 = “strongly agree.”
Gender norms
The Gender-Blind Sexism Inventory (Stoll et al., 2017) assesses individuals’ views on gender equality and captures both benevolent and hostile sexism. This measure, consisting of 11 items, was used to assess individual’s endorsement of gender norms. Prior work has provided evidence of the scale’s validity (Stoll et al., 2017) and reliability (Cronbach’s α = .80; Stoll et al., 2017), and it also demonstrated good reliability in the present study (Cronbach’s α = .88). An example item is, “affirmative action policies benefit women at the expense of men” (Stoll et al., 2017, p. 35). Participants were asked to rate their agreement of the 11 statements using a 5-point Likert scale, ranging from 1 = “strongly disagree” to 5 = “strongly agree.”
Bystander attitudes
Participants also completed the Bystander Attitude Scale (Banyard et al., 2005), which assesses bystander attitudes by asking participants how likely they are to intervene as a bystander in a variety of ways. This measure, consisting of 51 items, allowed us to assess if bystander behavior and attitudes align by comparing an intervention behavior in our vignette and a self-reported measure of bystander attitudes. Prior work has provided evidence of the scale’s validity (Banyard et al., 2005) and reliability (Cronbach’s α = .90; Banyard et al., 2005), and it also demonstrated good reliability in the present study (Cronbach’s α = .99). An example item is, “try to get help if I suspect a stranger at a party has been drugged” (Banyard et al., 2005, p. 14, Supplemental Appendix D). Participants were asked to rate their likelihood of engaging in the behaviors illustrated in the 51 statements using a 5-point Likert scale, ranging from 1 = “not at all likely” to 5 = “extremely likely.”
Demographics
Finally, participants answered a series of questions concerning their demographics. These questions included basic demographic variables such as gender, race, ethnicity, age, and sexuality.
Additional measures
A number of additional exploratory measures were collected that were not analyzed as part of the present investigation, including: The Just World Scale (Rubin & Peplau, 1975) and The Interpersonal Trust Scale (Rotter, 1967).
Procedure
After informed consent, participants were instructed to put on their headphones and they listened to a demo audio clip of white noise to test that they could hear the audio stimuli being presented by the program. Following this, participants were randomly assigned to one of the four audio vignette conditions and received instructions for and completed the bystander task as described above. After the bystander task, participants completed a series of questionnaires as described above, and were offered an opportunity to listen to the full audio clip if desired. The entire protocol took around 20 to 25 minutes to complete. The participants were then given a written debriefing and a list of resources to national hotline services for mental health and sexual assault and were thanked for their participation.
Analysis Plan
Preliminary data screening
As noted above, 45 participants who consented to the study but completed less than half of it were excluded prior to analyses. Of the remaining 402 participants, 14 participants who answered both comprehension check questions incorrectly (3%) were also excluded from the sample prior to analyses. Seven participants did not report their race and 15 did not report their ethnicity. In addition, 10 individuals listed a gender (e.g. woman) for their sexual orientation and were excluded from analyses involving sexual orientation; there was no other missing data. Screening of continuous variables did not identify any potential outliers or violations of normality assumptions, except for response time data. Response time data was not normally distributed, and non-parametric analyses were used to analyze this variable. Further, screening did not reveal any concerns about multicollinearity for multiple regression analyses or heterogeneity of variance for analysis of variances (ANOVAs).
Approach
Comparisons across vignette conditions were made with one-way ANOVAs for normally distributed continuous variables, with Fisher’s least significant difference (LSD) tests for post hoc comparisons. A non-parametric independent-samples Kruskal–Wallis test was used to compare time to intervene across conditions as this variable was not normally-distributed. Comparisons across vignette conditions for categorical or binary variables were made with chi-square analyses with z-tests for post hoc comparisons of proportions across pairs of conditions. Associations between individual difference variables and intervention behavior (coded as 1 = intervened; 0 = did not intervene) were assessed using point-biserial correlations or chi-squares as appropriate. Finally, we explored whether any of the bystander characteristics assessed here were differentially related to bystander intervention behavior across the different vignette conditions using a series of binary logistic regressions with intervention behavior as the binary outcome variable. A separate binary logistic regression model was specified for each of the six bystander characteristics, with condition (dummy-coded), bystander characteristic, and their interaction terms included as simultaneous predictors in the model. Corrections for multiple comparisons were not applied given that this is exploratory as opposed to confirmatory research, in line with published recommendations (Bender & Lange, 2001; Streiner & Norman, 2011). As with any preliminary study, further confirmatory studies are needed to test the reliability of findings (Nosek & Lakens, 2014; Wagenmakers et al., 2012).
Results
Perceptions of the Vignettes
Means and standard deviations for the six questions evaluating the vignettes are provided in Table 2, for the sample as a whole as well as for each vignette condition separately. A series of one-way ANOVAs failed to reveal significant differences across vignette conditions for ratings of whether the interaction would be considered sexual assault, F(3, 384) = 1.86, p = .137, η2 = .014, 95% CI [0.00, 0.04], as well as ratings of the perceived realism of the interaction, F(3, 384) = 1.55, p = .201, η2 = .012, 95% CI [0.00, 0.04], though ratings were generally high, indicating the vignettes were perceived as depicting sexual assault (M = 5.79, SD = 1.35) and as sufficiently realistic (M = 5.54, SD = 1.14). Ratings of the perceived likelihood that another person would behave in a manner similar to the aggressor were fairly high on average (M = 5.31, SD = 1.30), and did not differ significantly across conditions, F(3, 384) = 1.34, p = .262, η2 = .010, 95% CI [0.00, 0.03]. Ratings of the perceived likelihood that another person would behave in a manner similar to the victim were also fairly high on average (M = 5.64, SD = 1.19), though these ratings did differ significantly across conditions, F(3, 384) = 2.73, p = .044, η2 = .021, 95% CI [0.00, 0.05]. Specifically, participants viewed the behavior of the victim as less likely in the female aggressor, male victim vignette (M = 5.36, SD = 1.39) compared to in the other three vignettes, for which ratings did not differ (male aggressor, female victim [M = 5.70, SD = 1.13]; male aggressor, male victim [M = 5.68, SD = 1.11]; female aggressor, female victim [M = 5.83, SD = 1.07]) (see Table 2). In addition, participants identified the gender of the victim as consistent with the gender of the victim intended by the vignette approximately 95% of the time, and the gender of the aggressor as consistent with the gender of the aggressor intended by the vignette approximately 94% of the time. Taken together, results suggest that the audio vignettes were fairly balanced across conditions; participants found all of the vignettes to be fairly realistic, thought they depicted an unfolding incident of sexual assault, and readily perceived the intended gender of both the aggressor and victim across vignettes.
Perceptions of the Vignettes: for Full Sample and by Condition.
Note. agg = aggressor; vic = victim.
Likelihood of Intervening by Condition
Of the total analytic sample (N = 388), 74% of the participants intervened during the bystander task (M = 0.74, SD = 0.44; see Figure 1A). A chi-square analysis revealed that condition was not significantly associated with likelihood of intervening, χ2(3) = 4.98, p = .173, η = .113, Bootstrapped 95% CI [0.05, 0.23]. However, post hoc comparisons of proportions across pairs of conditions using z-tests revealed that participants in the Male Aggressor, Female Victim condition were significantly more likely to intervene (81.4%) than were participants in the Male Aggressor, Male Victim condition (67.7%), p < .05, with the likelihood of intervening in the other two conditions with a female aggressor falling intermediate (Female Aggressor, Female Victim condition = 75%, and Female Aggressor, Male Victim condition = 72.9%; see Figure 1A).

Bystander intervention behavior: for full sample and by condition.
Time to Intervene by Condition
As depicted in Figure 1B, participants intervened on average in 134.69 seconds (SD = 53.47), approximately 56% of the way through the 4-minute vignettes (response time for individuals who did not intervene was set to the length of the vignette). An independent-samples Kruskal–Wallis test revealed there were not significant differences in time to intervene across conditions, H(3) = 4.58, p = .206, η2 = .004 (see Figure 1B). Exploratory analyses involving time to intervene largely failed to reveal any significant effects, including associations with the bystander characteristics and attitudes measured. This same pattern of null results was observed when individuals who did not intervene were omitted from response time analyses. For the sake of brevity, we do not report these null results here, though interested readers are able to explore the reaction time data, which is available online (https://osf.io/h2dfb/).
Individual Differences in Likelihood of Intervening
Using point-biserial correlations, analyses revealed participants were significantly more likely to intervene in the audio vignette if they self-reported greater intentions to intervene on the bystander attitude survey, r(388) = .19, p < .001, 95% CI [0.09, 0.29]; lesser rape myth acceptance, r(388) = −.19, p < .001, 95% CI [−0.28, −0.07]; or lesser gender-blind sexism, r(388) = −.18, p < .001, 95% CI [−0.28, −0.09]. Self-reported intentions to intervene on the bystander attitude survey were similarly significantly associated with lesser rape myth acceptance, r(388) = −.39, p < .001, 95% CI [−0.49, −0.28] and lesser gender blind sexism, r(388) = −.46, p < .001, 95% CI [0.54, −0.37], which were strongly positively associated with one another, r(388) = .78, p < .001, 95% CI [0.74, 0.81].
In addition, in an analysis involving only participants who identified their gender as woman/female or man/male (n = 367), women were more likely to intervene (76.4%) than were men (69.0%), though this comparison failed to reach significance, χ2 = 2.49, p = .115, η = .082, bootstrapped 95% CI [0.01, 0.19]. As predicted, those identifying as LGBQ+ were significantly more likely to intervene (83.1%) than were those identifying as heterosexual (66.5%), χ2 = 13.68, p < .001, η = .190, bootstrapped 95% CI [0.09, 0.29]. Contrary to predictions, participants with at least one prior experience involving unwanted sexual activity (including unwanted sexual contact or intercourse) were not significantly more likely to intervene (74.6%) than were those with no prior experiences of unwanted sexual contact (73.5%), χ2 = .06, p = .810, η = .012, bootstrapped 95% CI [0.00, 0.11].
Interactions Between Individual Differences and Vignette Condition
Finally, we explored whether any of the bystander characteristics assessed here were differentially related to bystander intervention behavior across the different vignette conditions using binary logistic regressions with whether the person intervened (coded 1 = intervened; 0 = did not intervene) as the outcome variable. A separate binary logistic regression model was specified for each of the six bystander characteristics, with condition (dummy-coded), bystander characteristic, and their interaction included as simultaneous predictors in the model. These analyses failed to reveal significant interactions between vignette condition and any of the six bystander characteristics (i.e. bystander attitudes, rape myth acceptance, gender-blind sexism, bystander gender, and bystander sexual orientation) (Table 3). This pattern of results suggests that the associations observed between these bystander characteristics and intervention behavior were more or less consistent across the different vignettes.
Significance Tests of Interactions Between Bystander Characteristics and Vignette Condition on Likelihood of Intervening.
While the overall interaction for vignette condition and having had a prior unwanted sexual experience was also not significant, Wald’s χ2(3) = 7.53, p = .056, examination of the individual model coefficients revealed a significant interaction coefficient for comparisons between the male aggressor/male victim and male aggressor/female victim conditions, Wald’s χ2(1) = 6.53, p = .011, OR = 7.94, 95% CI [1.62, 38.96] (see Figure 2). Separate binary logistic regressions for individuals with and without at least one prior unwanted sexual experience revealed that while individuals with at least one prior unwanted sexual experience demonstrated relatively stable rates of intervening across all vignette conditions, Wald’s χ2(3) = .22, p = .975, individuals without any prior unwanted sexual experiences demonstrated significant differences in the odds of intervening across vignettes, Wald’s χ2(3) = 11.09, p = .011. Specifically, individuals without any prior unwanted sexual experiences were significantly less likely to intervene in the male aggressor, male victim condition compared to the male aggressor, female victim condition, Wald’s χ2(1) = 9.63, p = .002, OR = 0.11, 95% CI [0.03, 0.45].

Interaction between vignette condition and prior unwanted sexual experience on intervening behavior.
Discussion
Consistent with predictions, results revealed that individuals were more likely to intervene in an unfolding incident of IPV if they were LGBQ+ (vs. heterosexual), or if they self-reported less rape myth acceptance, greater bystander intentions, or more gender equality endorsement. Contrary to predictions, we failed to find any significant differences in intervention behavior based on prior unwanted sexual activity or bystander gender, although this finding was in the expected direction. We also found that, in general, individuals were significantly more likely to intervene in the vignette condition most aligned with stereotypical gender roles (i.e. the vignette with a heterosexual couple with a male aggressor and female victim) compared to all other vignette conditions, though this comparison only reached significance with the male aggressor, male victim condition.
The findings of the current study are largely consistent with past literature on the topic of IPV and bystander intervention. Prior literature has supported the hypothesis that LGBQ+ identifying individuals may be more likely than heterosexual individuals to intervene in a developing incident of IPV (Edwards et al., 2015; Hoxmeier et al., 2022), and this pattern was also identified in our study. Similarly, our findings show that individuals who report less rape myth acceptance and/or more gender equality endorsement are more likely to intervene is also consistent with previous findings (Banyard & Moynihan, 2011; Davies, 2002; Davies et al., 2012; McMahon, 2010). Contrary to prior literature, however, our study did not find significant differences in intervention behavior based on the gender of the bystander. Previous studies have found relationships between gender identity and intervention behavior such that those who identify as women are more likely than men to intervene in an instance of IPV (Bennett et al., 2017; Seelau et al., 2003).One interesting possibility for this null result is that there are gender differences in the intention to intervene but not in actual intervention behavior, as past work has relied largely on self-reported intervention intentions, whereas the present study measured actual intervention behavior using an audio vignette. Future research should probe these possible gender differences in IPV intervention behaviors and intentions more directly.
Critically, we observed few significant interactions between vignette condition and bystander characteristics in predicting intervening behavior, suggesting bystander characteristics may contribute to bystander intervention behavior consistently across incidents of IPV that vary in terms of the apparent gender and sexual orientation of the individuals involved. However, like all null effects, these null interaction effects should be interpreted with caution; a number of things could contribute to the lack of differences found. For one, the vignette script was identical across all conditions, and it was developed based on a script from a past study (Marx & Gross, 1995) using a heterosexual couple with a male aggressor and female victim, the most stereotypical and heteronormative relationship and IPV scenario. Keeping the script and context the same across conditions helped us to isolate any potential causal impact of the apparent gender and sexual orientation of the individuals in the vignette on bystander behavior, but this also means that the vignettes do not capture potentially important differences in how instances of IPV may unfold across various types of relationships in the real world. An important direction for future work is to develop and incorporate scripts built from these perspectives as well.
Although the bystander characteristics that influenced intervening behavior largely failed to interact significantly with vignette condition, there was a significant interaction between condition and having had a prior unwanted sexual experience to predict bystander behavior: individuals with at least one prior unwanted sexual experience intervened at a fairly equivalent rate across vignette conditions, whereas those with no prior unwanted sexual experiences were most likely to intervene in the male aggressor, female victim condition and least likely to intervene in the male aggressor, male victim condition. This pattern of results suggests that, compared to those who have experienced at least one incident of unwanted sexual activity in their past, those without any history of experiencing unwanted sexual activity are altering their intervention behavior more based on the apparent gender and/or sexual orientation of the individuals involved in an unfolding incident of IPV. This is consistent with hypothesizing that a prior unwanted sexual experience may lead to increased empathy for victims, enabling those with a prior experience of unwanted sexual activity to overcome situational factors (e.g. the gender or sexual orientation of the victim) that might otherwise make a person less likely to intervene in an incident of IPV (Leone et al., 2021). Additionally, having had an experience involving unwanted sexual activity may be associated with greater IPV efficacy (McMahon et al., 2020), such that a better understanding of IPV warning signs may lead a bystander to recognize an IPV event more readily or confidently. It should be noted that we did not observe any significant differences in time to intervene based on prior experiences with unwanted sexual activity in the present study, suggesting those with prior experiences with unwanted sexual activity were not necessarily faster to recognize the situation as needing intervention in our particular audio vignette. Still, this idea that prior experience may increase efficacy could help to explain why individuals may be most likely to intervene in the male aggressor, female victim condition when they have not experienced unwanted sexual activity in the past themselves, as it is the most stereotypical and therefore recognizable instance of IPV.
This study adds to the growing body of literature on bystander behavior in IPV using an ecologically valid task where participants were asked to listen and respond to an audio vignette of an instance of IPV unfolding in real time. This methodology has a number of advantages over more traditionally utilized techniques, such as having participants self-report whether or not they think they would intervene in hypothetical scenarios, which measured intended as opposed to actual intervention behavior and are likely less evocative than audio vignettes. In addition, many existing studies focus solely on the existence of IPV in heteronormative relationships and in relationships with male-led aggression, but the present study adds to the limited knowledge concerning reactions to IPV in a wider variety of relationships and relationship dynamics, such as in LGBQ+ identifying relationships or relationships with female-led aggression. Subsequently, the present study also contributes to current knowledge regarding less frequently studied bystander characteristics that may contribute to one’s likelihood of intervening during an observed instance of IPV, such as sexual orientation and societal beliefs.
Limitations and Future Directions
This study used an online platform to recruit participants and collect data, so the results are subject to biases stemming from under-coverage and nonresponse. Not everyone in the United States has access to the Internet, and there are demographic differences between those who do and do not. There may be attributes of lower income groups, individuals in rural areas, or older individuals without internet access or technological savvy that may make them more or less likely to experience IPV, but they will have been left out of this study. Similarly, the study featured a majority White and cisgender participants, and therefore it is unclear how well the results generalize to individuals of other racial or ethnic groups, or individuals with other gender identities. These are critical areas of exploration for future research. Relatedly, our sample size (N = 388) would be underpowered for examining the interactive and/or compounding influence of multiple identities on bystander behavior across vignettes, particularly for groups that are not in the majority in the population. Though note that in our sample nearly half of participants identified as LGBQ+ (48.5%) and more than half reported experiencing some form of unwanted sexual experience (66.0%), indicating we had fairly strong representation of individuals from these historically under-represented groups. Lastly, the study utilized several self-report questionnaires, and this type of data collection can allow for biased or untruthful answers to be introduced, as some participants may feel uneasy about answering potentially controversial questions.
This study could also serve as a starting point for developing novel IPV prevention programs in the future that incorporate examples of IPV in non-heterosexual relationships or relationships with female led aggression. Although the rates of violence for LGBQ+ communities are higher than those for the rest of the population (Edwards et al., 2015; National Coalition of Anti-Violence Programs, 2016), prevention programming and resources that factor in marginalized identities are lacking (McMahon et al., 2020). With education comes greater awareness of and efficacy in identifying the warning signs of IPV across varying relationship dynamics. Prevention education or training that highlights a broader spectrum of identities may help bystanders to identify and feel more comfortable intervening in less heteronormative instances of IPV, including those involving same-sex relationships or relationships with female-led aggression. Finally, in light of the observed associations between IPV bystander intervention and both rape myth acceptance and gender equality beliefs, such factors should be targeted by future intervention training programs in order to increase rates of bystander helping behavior.
Supplemental Material
sj-docx-1-jiv-10.1177_08862605221111412 – Supplemental material for Bystander Intervention in Intimate Partner Violence: An Audio Vignette Study of Heterosexual, Gay, and Lesbian Dating Partners
Supplemental material, sj-docx-1-jiv-10.1177_08862605221111412 for Bystander Intervention in Intimate Partner Violence: An Audio Vignette Study of Heterosexual, Gay, and Lesbian Dating Partners by Kristen Graham, Abby LaChance and Jolie B. Wormwood in Journal of Interpersonal Violence
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
