Introduction

Criminological research has traditionally regarded intimate partner violence (IPV) as an epidemic of violence against women perpetrated by men in the context of heterosexual relationships (Felson, 2002; Messinger, 2011). As has been noted, “when most people think of a victim of partner abuse, they think of a heterosexual woman” (Russell et al., 2015, p. 48). Same-sex partner violence has only recently begun to be included in the core conceptualization of research on IPV (Baker et al., 2013; Hamby, 2009). This work increasingly finds that the prevalence and severity of physical violence among sexual minority couples are similar to and, in many cases, higher than among heterosexuals (Ard & Makadon, 2011; Bender & Lauritsen, 2021; Blosnich & Bossarte, 2009; Breiding et al., 2013; Dank et al., 2014; Graham et al., 2019; Tjaden & Thoennes, 2000). Same-sex victims of IPV, however, are especially reluctant to seek assistance from the police and consequently are often left unprotected from the law, and their perpetrators go unpunished (see Briones-Robinson et al., 2016; Girardi, 2022). This special unwillingness of same-sex victims to enlist the criminal justice system underscores the critical role of third parties as agents of legal mobilization.

An important question, therefore, is whether the public—third-party witnesses—regard partner violence between same-sex and opposite-sex couples as equally deserving of criminal justice intervention? The literature on the normative and social psychological basis of biases and decision-making offers answers to this question. On one hand, victims of same-sex partner violence may attract less third-party support because they do not conform to traditional gender role expectations and are possibly subject to discriminatory treatment (Harris & Cook, 1994). According to Russell et al., (2015, p. 49), sexual minority victims “endure heterosexist attitudes” from public officials and the wider community. On the other hand, perhaps it is the gender of disputants rather than their sexual orientation that determines when third parties mobilize the criminal justice system. Physical power imbalances between men and women and a chivalry norm combine to prioritize the protection of women (versus men) from harm (Felson, 2002; Harris & Cook, 1994; Rogers et al., 2019). In addition to the foregoing assumptions, the severity of violence theoretically moderates the effects of gender and sexuality on third-party reactions to partner violence (see Felson & Cares, 2005). Severe cases of partner violence theoretically elicit strong third-party responses regardless of the sexuality or gender of disputants because they heighten the perceived costs of inaction (Kidd, 1979). In this way, the harm suffered by victims potentially tempers the effects of disputants’ attributes (i.e., gender and sexuality) on third-party responses.

Only a limited number of criminological studies have examined third-party legal mobilization in response to witnessing partner violence between same-sex and opposite-sex couples. This dearth of research reflects what Cannon and colleagues (2015, p. 68) attribute to a “heteronormative bias that runs throughout most domestic violence scholarship.” Generally, however, the empirical evidence is mixed and conflicting, much of which is derived from college-based convenience samples (e.g., Russell et al., 2015; Seelau et al., 2003). More broadly, foundational work on perceptions of same-sex violence (e.g., Harris & Cook, 1994; Poorman et al., 2003; Seelau et al., 2003) originates from an earlier era when public attitudes towards gender minorities were less accepting than today.

Experimental vignettes can effectively isolate the comparative effects of gender, sexuality, and incident severity on respondent intentions while controlling for incident characteristics (e.g., provocations, location). Given the low prevalence of gender minorities, paired with the infrequency of serious violence in observational data, experimental vignette designs provide a sound and efficient alternative for theoretical testing (Horne & Lovaglia, 2008). Moreover, such designs effectively standardize contextual factors that are likely to guide human decision-making. Using a factorial vignette administered to an M-Turk online sample of 803 adults, this study examines whether either the gender or sexuality of disputants affects third-party legal mobilization.

Examining these aims is important for several reasons in the broader theoretical context of viewing law as a dependent variable (e.g., Black, 1976). Citizen reporting is the primary mechanism through which violent crimes become known to law enforcement (Berg & Rogers, 2017). When third parties are less willing to report violence perpetrated against certain victims, this hinders the effective deployment of police resources and can bias official crime data, contributing to the dark figure of crime. Such decisions ultimately affect the allocation of crime prevention resources and whether victims receive equal protection under the law regardless of their gender or sexuality (Seelau & Seelau, 2005). Moreover, the absence of formal law raises the possibility that victims will rely on extralegal forms of self-help to deter their attackers and seek retribution (Black, 1976). Determining how the sexuality and gender of disputants correlates with variability in legal mobilization will extend the broader literature on criminal legal responses to partner violence.

Background

Historically, various sociolegal perspectives in criminology and sociology have been developed to articulate the normative, relational, and social psychological factors that predict variation in legal mobilization outcomes. How third parties or bystanders respond to violent disputes varies with the attributes of victims, privacy norms, and the seriousness of the situation (Berg & Rogers, 2017). For instance, Black’s (1976) theory treats law as a dependent variable whose quantity varies with the position of individuals across realms of social life. Decisions to mobilize the criminal law vary directly with what Black (1976) conceptualizes as the vertical, cultural, and radial of dimensions (see Xie & Lauritsen, 2012). Notably, Black’s theory is not concerned with the psychological states of disputants (e.g., biases, stereotypes), but only with their locations in social space. Kidd’s (1979) model of legal mobilization, however, relies heavily on social psychological and rational choice principles. This model argues that discrepancies between a bystander’s norms about correct behavior and observations of a crime theoretically cause cognitive dissonance in situations where bystanders feel responsible to act. According to Kidd (1979), reporting criminal activity to the police alleviates the arousal created by these discrepancies. Mobilizing the law brings material and psychological benefits to the bystander, whereas non-intervention can result in high psychological costs (e.g., self-image, social standing). Several psychological processes theoretically mitigate the likelihood of bystander reporting include the denial of harm to the victim, the unsuitability of the act for punishment, and sympathizing with the violator (Kidd, 1979). For some bystanders, non-reporting is the result of not attributing criminality to an act even if it is non-normative. More recently, Weiss (2021) argued that witnesses justify their decisions to not report crimes using the logic of “accounts”. By denying the seriousness of crimes and their own responsibility to act, bystanders justify and excuse their inaction on normative grounds.

Altogether, sociological theories of mobilization emphasize the importance of understanding incident level variation in decisions to mobilize the law on behalf of victims. While these models were not explicitly designed to assess the role of sexuality and gender, they provide a lens through which to conceive of how the characteristics of not only crimes but also victims theoretically affect case level variation in bystander reactions.

An interdisciplinary body of empirical literature offers explicit assumptions about how the sexuality of couples affects the intentions of third parties to mobilize legal intervention in response to witnessing intimate partner violence. One view suggests a prevailing homophobic culture in society tends to subject gay men and lesbian women to formal and informal types of discrimination (e.g., Hall & Rodgers, 2019; Herek, 2000). The unwillingness of citizens to aid lesbian and gay victims therefore reflects the continued devaluation of sexual minorities (Sorenson & Thomas, 2009). Some older studies of help-seeking behaviors find that the public offers less assistance to homosexual- compared to heterosexual individuals (Gore et al., 1997; Hendren & Blank, 2009; Shaw et al., 1994). The logic of Black’s (1976) basic model, while clearly silent about law and sexuality, may be extended to account for these patterns. In his model, same-sex persons would, in theory, possess lower cultural status (e.g., conventionality) relative to heterosexual couples. Furthermore, because law theoretically increases with social status, a grievance between heterosexual couples would attract a greater quantity of legal mobilization (law) from bystanders than disputes between homosexual couples.

Relatedly a line research suggests the unfair treatment of gay and lesbian persons stems “not only from their sexual orientation per se” but their “violations of traditional gender roles” (Blashill & Powlishta, 2009, p. 783). Sexual orientation thus serves as a shorthand cue of gender atypicality or gender role inconsistency. Women and men who fail to conform to traditional gender role expectations are deemed deviant, and their suffering is devalued (Stanziani et al., 2020). As Brown (2008, p. 461) reasons, “homophobia/heterosexism along with ignorance regarding intimate partner abuse and same sex couples creates barriers to reporting incidents of same sex violence.” Lesbian and gay victims of partner violence cite concerns about discriminatory treatment as a reason for not seeking assistance from outsiders (Balsam & Szymanski, 2005; Callan et al., 2021). Consistent with these views, respondents are more likely to perceive IPV assaults as abuse in opposite sex compared to same-sex couples (Russell et al., 2015). Studies have also found that the police minimize the seriousness of partner violence incidents when the disputants were in a same-sex relationship (Comstock, 1991). Perhaps bystanders do not impute the same level of criminality or harm to acts of violence between same and opposite-sex couples.

Gender stereotypes, sexual minorities, and third-party legal mobilization

A consideration of disputant gender is necessary to formulate predictions about how the public reacts to homo- and heterosexual partner violence (Seelau & Seelau, 2005). Stereotypes of the male and female sex typically attribute the masculine qualities of aggressiveness and dominance to men, conceiving of women as the more vulnerable, weaker gender (Pierce & Harris, 1993). Through these gendered-lenses, the victim’s role (e.g., passivity, helplessness) is more often viewed as a “feminine trait” whereas male-centered masculinity aligns with the offender role (Dardis et al., 2017).

Most but certainly not all of what is known about how disputant gender shapes third-party reactions to IPV is based on studies of heterosexual couples. This work finds that man-on-women violence is rated as more illegal than woman-on-man violence (Sorenson & Taylor, 2005). Survey respondents also favor harsher penalties for aggressors when their victims are women. For instance, men accused of IPV are viewed more negatively than women by mock jurors (Stanziani et al., 2018), and defendants convicted of violent crime against women (versus men) tend to receive harsher sentences in criminal courts (Stauffer et al., 2006). Other research shows that men are more likely to be found guilty by mock jurors than women for the same crime (Kern et al., 2007). Respondents view male perpetrators as less likeable than female perpetrators in scenarios of couple violence (Harris & Cook, 1994). Similarly, the public expresses a greater desire to protect female rather than male victims of partner violence (Poorman et al., 2003). As Brown notes (2008, p. 460), “one of the biggest impacts of gender role socialization on intimate partner violence is the belief that only females can be victims and only males can be perpetrators.” Overall, then, third-party evaluations of partner violence tend to be more punitive towards males who victimize women rather than the reverse.

Some scholars assert that third parties are more inclined to support women versus male victims due to physical power imbalances between men and women (Archer, 2004; Eagly & Crowley, 1986; Felson, 2002). If so, norms protecting women from harm reflect a general tendency to protect individuals—regardless of their gender—who are perceived to be physically vulnerable (see Berkowitz & Daniels, 1963; Fold & Robinson, 1998). This vulnerability norm expects violence against women (as opposed to men) to be more strongly condemned due to societal norms which value the defense of weaker individuals. For example, man-on-woman violence is rated more frightening than woman-on-man violence because of gender imbalances in strength (Hamby & Jackson, 2010). Seelau and Seelau (2005, p. 369) argue that female victims are seen as more vulnerable than men, in part, due to perceived estimates of discrepancies in the potential for harm. Likewise, third parties are more likely to aid women victims than men in violent conflicts due to perceived male strength advantages (Rogers et al., 2019). The research on gender role stereotypes (e.g., Sorenson & Taylor, 2005) and the vulnerability norm (e.g., Fold & Robinson, 1998) therefore suggests that heterosexual incidents of violence involving female victims will generate the strongest third-party reactions. That is to say, the typical heteronormative conceptualization of intimate partner violence—in which men attack women—may be considered the worthiest of legal intervention compared to other heterosexual and homosexual relationship configurations (Stanziani et al., 2018).

Whether the gender of disputants maximizes or minimizes third-party intentions to mobilize the law in response to same-sex relationships is not clear in the empirical literature. The foregoing theoretical literature on the vulnerability norm suggests that violence in same-sex relationships would be perceived as the least deserving of legal intervention because disputants are believed to be of similar size and strength (Russell et al., 2015). Furthermore, the gender role socialization research would suggest that violence in lesbian relationships may be considered worthier of intervention than violence against men in gay relationships. Norms of chivalry might also attract more support for lesbian women compared to gay male victims. Conversely, the literature suggests that third parties may be less likely to mobilize the law for male victims of same sex compared to heterosexual violence (Russell et al., 2015; Sorenson & Thomas, 2009). Not only do gay male victims violate social norms of masculinity, but they also do not fit heteronormative perceptions of intimate partner violence (Stanziani et al., 2018)—meaning, gay men are not as likely to be considered worthy of protection (Russell et al., 2015).

A small line of research regarding these predictions is mixed and inconclusive: some studies demonstrate that violence by heterosexual men against women elicits stronger reactions than acts committed by heterosexual women and same-sex partners (Ahmed et al., 2013; Russell et al., 2015), whereas other work finds partial support for the effects of couple sexuality on third-party action (Stanziani et al., 2018). For instance, a vignette study found that police officers rated heterosexual male perpetrators of partner violence as more of a threat compared to gay male, lesbian, and heterosexual female partners (Russell et al., 2015; Seelau et al., 2003). Still, vignette research has found that intimate partner violence between two men was the least likely to be considered abuse by respondents (Russell et al., 2015). According to a mock jury study, violence perpetrated against men (by male or female partners) and by women against persons of either gender was of a “lesser caliber” compared to other couple types (Stanziani et al., 2018). Generally, but not always, the sex of the victim, regardless of their sexuality, is among the strongest predictors of third-party intentions to enlist the criminal justice system (Seelau & Seelau, 2005; Sorenson & Thomas, 2009).

Severity of intimate partner violence incidents

There are good theoretical reasons to suggest that the severity of violence between intimate partners conditions the effects of sexuality and gender on third-party responses. The degree of physical harm felt by victims affects the blame assigned to perpetrators and whether witnesses feel responsible for punishing them (Felson, 2002). Moreover, the degree of physical harm increases the perceived moral and normative costs to third parties of not rendering assistance. According to Kidd (1979), failing to mobilize the law in response to especially severe incidents of crime can cause bystanders to incur high psychological costs. Studies have shown that the severity of violence between intimates—assessed by the potential or actual physical harm—is an important determinant of whether third-party observers believe incidents are worthy of legal intervention (see Berg & Rogers, 2017; Gracia & Herrero, 2006). For instance, perpetrators that use severe forms of violence against their partners tend to receive harsher criminal punishments (Lane & Knowles, 2000). Moreover, the severity of violent incidents significantly predicts the willingness of actors to report incidents to the police (Gracia & Herrero, 2006; Gracia et al., 2009). This work generally suggests the intolerability of violence motivates third-party actions. This reaction, however, may depend on the gender of disputants. The media presents the same acts of violence differently for male and female perpetrators with severe violence committed by men attracting greater media attention (Carlyle et al., 2008).

The literature suggests two potential effects of violence severity on third-party reactions to same-sex versus opposite-sex violence: For one, perhaps severe forms of partner violence may reduce the effects of disputant sexuality and gender on the intentions of third parties to mobilize the law. That is, partner violence between men or between women might generate comparable third-party attention only when the potential for harm is at a maximum—when violence tends to be most condemned. It may be that the public becomes increasingly egalitarian in instances with a high potential for physical harm. The intolerability of severe violence might override norms prioritizing women and heterosexual victims. Alternatively, perhaps third parties are just as biased towards heterosexual and women victims (versus male and homosexual victims) regardless of severity. Kidd’s model (1979: 391) suggests bystanders might discount the significance of a crime because it is not salient to their normative beliefs about what is and what is not acceptable behavior. All of which allows them to mitigate mobilizing the law even in serious incidents of violence. Here, violence between homosexual intimates might not cue salient normative concerns for certain actors. If so, and if severity does not equalize reactions to IPV among same versus opposite-sex couples, it would support the assumption that the public is biased against non-traditional victims of partner violence.

Current study

The current study applies an experimental design to assess the willingness of third parties to mobilize the law in two related ways: (1) to notify the police and (2) to punish the perpetrators with a jail sentence. The study examines three questions: First, are third parties less likely to mobilize the law in response to witnessing violence between same-sex vs. opposite-sex couples? Second, to what extent are third parties more responsive to the gender of intimate partner disputants as opposed to their sexuality? The latter question addresses whether gender is a stronger determinant of legal intervention than the sexuality of disputants. Third, to what extent does the severity of violence moderate the impact of sexuality and gender on the willingness of third parties to mobilize the law?

Methods

Participants and procedure

To address the research questions, the study employed an experimental design with vignettes to examine respondents’ likelihood of mobilizing the law in response to witnessing episodes of partner violence among gay, lesbian, and straight couples. The pilot phase of the study was conducted at a large Midwestern university with undergraduate students. Selection procedures were based on convenience as students who participated were selected from an introductory sociology course. A total of 71 students participated in the pilot study. The main objective of the pilot was to test the data collection instruments with a particular focus on the length of the survey, clarity, and realism of the scenarios described in the vignettes, phrases, or terms that were redundant or confusing among other facets of the design. Because this pilot was concerned with instrumentation, the University’s Institutional Review Board deemed the pilot exempt and did not allow for demographic information to be collected from respondents.

The undergraduate students were asked to provide feedback about changes that should be made to the survey. The feedback included adding a back button on the survey to return to previous questions, presenting the reporting and punishment decision questions on the same page as the vignette, and clearly stating with direct phrasing that the couples are married. Using this feedback, data collection tool changes were made prior to the current study’s development.

Also using the pilot study responses, a power analysis was conducted to determine the appropriate sample size for the study design. According to the analysis, to achieve a power of 0.85, the minimum sample size needed was 70 subjects per condition or 560 participants total. Subjects in this study were recruited from Amazon Mechanical Turk (MTurk) in February 2020. MTurk is an online platform that allows users to earn monetary credit for completing surveys and has become a widely used source of data for aggression and crime research in recent years (Overton et al., 2022; e.g., Seigfried-Spellar et al., 2017). As recently noted in Shi (2022, p. 282), studies have shown that MTurk samples can produce results “similar to those from nationally representative samples.” To be eligible for participation in the study, participants must have had an approval rating of at least 95% (see Peer et al., 2014), reside in the USA, and be at least 18 years of age. Respondents earned 20 cents upon completion of the survey via a Qualtrics platform. All respondents were residents of the USA and 18 years or older. A total of 821 MTurk workers participated in the study. Of these, 16 (1.9%) respondents had a response time less than 60 s (1 s per survey item) and were not included in the analysis (see Wood et al., 2017). Upon further inspection, 11 of the 16 with response times less than 1 s per item were full non-complete responses (missing data on each question). An additional two respondents were missing data on each question in the survey and were excluded from the analysis.

Following the exclusion of the 18 responses that were non-complete and/or had response times less than 1 s per item, an analysis of missing data revealed that 19 respondents were missing data on key variables in the model. To address the missing values, multiple imputation using chained equations (MICE) was utilized to allow the specification to be fully conditional (Van Buuren et al., 1999). All variables included in the analysis were included in the MI models, six of which contained missing values. For context, variables with missing data were self-reported violent acts (n = 8), whether the respondent had ever been arrested (n = 7), agreement with gay marriage rights (n = 7), race (n = 9), whether the respondent is involved in a committed relationship (n = 14), and level of education (n = 11). The imputation was implemented using the mi impute chained command in Stata version 18, with the default number of 10 cycles (Royston, 2004). This step resulted in a final sample size of 803 respondents. The data and Stata code used for the analysis are available upon request from the first author.

Vignette and independent variables

Participants in the study were first asked to provide their demographic information and to answer a series of survey questions pertaining to their beliefs and past experiences with matters relevant to their intentions to engage the law (Table 1). Respondents then read a vignette describing an act of violence between two intimates and were prompted to report how they would respond to witnessing the act. The design included eight conditions that varied the gender of the aggressor (male/female), the gender of the victim (male/female), and the severity of the violent act (“punch in face and knocked to ground”/ “slap across face”). Note that the design varies the sexuality of the couple by varying the gender of the disputants. The design relies on a between-subjects comparison for which respondents were assigned one of eight possible experimental conditions.

Table 1 Descriptive statistics for selected variables included in analysis (N = 803)
Full size table

As noted, how third parties both interpret and respond to acts of violence can be affected by the social context of the violent incident. Situational factors are strong determinants of police notification patterns (Rennison, 2010). It is important that the scenarios described in the vignettes be the same in every respect except for the focal variables to minimize differences between the incidents. For this reason, the vignette held constant features of social context and the relationship between the respondent and the two disputants. Table 2 provides an overview of the distribution of each condition and independent variable in the study sample.

Table 2 Number of respondents per vignette condition in (N = 777)
Full size table

Below is the experimental vignette:

Imagine that you are walking alone through a public park. It is a quiet night, and the park is empty. As you are walking, you notice [two men/a man and woman/two women] across that park that you recognize from around your neighborhood. You know that these two are married to each other because you have seen them holding hands in public and you’ve noticed that they both wear wedding bands. All of the sudden, you witness [one of the men/the woman/the man/one of the women] get very angry at [his husband/his wife/her husband/her wife]. The angry [man\woman] swears at [his husband/his wife/her husband/her wife] and slaps [him/her] across the face—OR—The angry [man/woman] swears at [his husband/his wife/her husband/her wife] and punches [him/her] in the face hard enough to knock [him/her] to the ground.

Outcome measures

After reading the vignette, respondents answered two questions gauging their intent to mobilize the law, the answers to which were employed as outcome variables: (1) how likely respondents are to “Call the police to report the attack” and (2) whether the “Aggressor should be forced to serve jail time.” Each was scored on a four-point Likert scale. Response categories for the first question include “Unlikely, Somewhat Unlikely, Somewhat Likely, Likely” and the categories for the second question were “Disagree, Somewhat Disagree, Somewhat Agree, Agree.”

Demographic and control variables

Respondents were also asked questions to identify their demographic characteristics, sexual orientation, attitudes, and past experiences with the criminal justice system. These variables improve the strength of the multivariate estimates and permit assessments of the composition of the MTurk sample. Table 1 provides detailed information on these study variables.

Demographic and lifestyle factors

The models include categorical measures of each respondent’s race (Black, White, Asian, or other) and gender (female, male, non-binary), and if they identify as transgender (yes/no). Also included is a seven-point ordinal measure of their highest level of education (1 = less than high school/7 = doctorate). A binary measure captures whether respondents were currently in a committed relationship (yes/no). Respondents also indicated whether they were a member of the LGBTQ + (queer) community (yes/no).

Compared to the US population demographics, the sample used for the study was slightly whiter (+ 2.49%) and female (+ 5.84%) (Bureau, 2020). Participants in the study were also slightly more educated with 61.06% holding a bachelor’s degree or higher compared to 33.7% of the US population. Additionally, 12.08% of the sample identified as a member of the LGBTQ + community, which is a larger proportion than the national average (7.1%) reported by a Gallup poll (2022).

Attitudes and personal experiences

Two measures assess respondent’s past involvement in crime and violence: a binary indicator of whether they had ever been arrested by the police (yes/no) and an ordinal measure of the frequency with which they engaged in violent acts in the past year (0 = never, 6 = weekly) (see Table 1). The study also includes an ordinal measure of the extent to which respondents agree that gay and lesbian person should be legally allowed to marry (1 = disagree, 4 = strongly agree):

Results

Sexuality and legal mobilization

Due to the ordinal nature of the two outcome variables, the analyses are conducted using ordinal logistic regression (McCullagh, 1980). As such, we estimate two baseline ordinal logit models in Stata (version 18) assessing whether couple sexuality separately predicts the two outcome measures in the hypothesized fashion. The ordinal logit models were performed with maximum likelihood estimation. Table 3 displays the results from these models. Turning to the police notification outcome, the model indicates that third parties were significantly less willing to notify the police when the couple in a violent altercation was gay as opposed to heterosexual (β − 0.34; p < 0.05). However, estimates comparing effects for lesbian versus heterosexual couples were not statistically significant. The effect of severity on police notification of law enforcement is statistically significant (β 1.45; p < 0.01) with respondents indicating a higher likelihood of reporting when the incident of violence is more severe. Regarding the jail as a punishment outcome, the regression models provided very similar estimates. Specifically, when the couple is gay rather than heterosexual, third parties are less likely to favor jail as a criminal punishment (β − 0.51; p < 0.01). Furthermore, the comparison between lesbian and straight couples was not significantly different. When the incident of partner violence is more severe, third parties more strongly endorse jail as a punishment for the perpetrator (β 0.97; p < 0.01).

Table 3 Ordinal logistic regression coefficients predicting notifying police and jail as punishment by sexuality and severity
Full size table

Disputants’ gender and legal mobilization

The gender of the aggressor and victim is confounded when solely including the sexuality variable in the model. As noted, each condition theoretically affects how third parties respond to acts of partner violence. The negative effect for gay partners could reflect the devaluation of male victims and not necessarily gay victims. A second set of regression models partitions intimate partner couples to isolate gender and sexuality contrasts. Table 4 displays estimates from these models revealing the additive associations between victim and offender gender and police notification and criminal punishment preferences.

Table 4 Ordinal logistic regression coefficients predicting notifying police and jail as punishment
Full size table

Model A of Table 4 indicates that third parties are the most likely to notify police on behalf of female victims specifically when attacked by their male partners compared to other couple combinations. This finding is supported by significant estimates for the “man hit man” (β − 1.35; p < 0.01), “woman hit man” (β − 1.98; p < 0.01), and “woman hit woman” (β − 0.76; p < 0.01) comparisons. Third parties are the least willing to report incidents involving male heterosexual victims, followed by gay male victims. Male (particularly heterosexual) victims, therefore, attract less third-party support than women. Moreover, only small differences in the likelihood of notification are detected when the female victim is a lesbian compared to heterosexual. Thus, across relationship types, the strongest effect is obtained for female victims. Model A also reports that incident severity has a strong additive effect on the willingness to notify police (β 1.57; p < 0.01). Model B builds on Model A (see Table 4) by including the remaining array of study variables. The results do not substantively change. Third parties are more likely to mobilize the law on behalf of women hit by their male partners than any other gender combination. Heterosexual male victims receive the least support. It is worth noting that the third party’s gender was also a predictor of notification, with female third parties indicating more willingness to notify police than men (β 0.54; p < 0.01).

A test of parallel regression was not completed on Model B of Table 4 with the imputed missing data because the standard Stata (version 18) postestimation platform was not capable of providing parallel regression test estimates. Rather, a Brant test was conducted on the model prior to the multiple imputation. The parallel regression test of Model B of Table 4 shows that the ordinal regression model does not violate the parallel regression assumption (p > 0.20) (Brant, 1990).

Overall, the results of Models A and B do not provide clear evidence that third parties are influenced or biased by the sexuality of the couple. Rather, the gender of the victim appears to have the strongest effects on third-party action, regardless of sexuality: women garner the most third-party support. Similarly, third parties are the least likely to mobilize the police for male heterosexual victims of partner violence mainly because the perpetrators are women.

Model C of Table 4 displays results for the jail punishment outcome measure. According to the estimates, third parties are most likely to support a jail sentence for male aggressors of partner violence—including both gay and heterosexual men. But the effects are strongest when the victim is a woman in a heterosexual partnership. These findings essentially mirror those of police notification: third parties are least likely to support a jail sentence for a female aggressor when the victim is male and are most likely to support a jail as punishment when the victim is female, and the perpetrator is a man. Moreover, regardless of who attacks the woman, third parties are more likely to mobilize the law on their behalf compared to incidents involving male victims.

Model D builds on Model C (see Table 4) by including the demographic and control variables of the study. While holding all else constant, the couple composition/role variables remain significant predictors of third-party support for jail as punishment. Third parties are more likely to mobilize the criminal legal system in incidents where women are hit by their male partners than any other combination. When victims are male—whether gay or heterosexual, third parties display weaker support for jail as a punishment for the offender. As before, female third parties indicating were more strongly in favor of jail sentences compared to male third parties (β 0.37; p < 0.01).

To test the parallel regression assumption, a Brant (1990) test was performed on Model D of Table 4 prior to the multiple imputation with chained equations (MICE). According to the results of the Brant test, the model does violate the parallel regression assumption (p > 0.004) (Brant, 1990). This means that the independent variables in the model do not have the same effect on the outcome variable for each cut point (level of agreement with jail as punishment). Specifically, the results of the Brant test indicate that the experimental condition of “severity” and whether the respondent is in a committed relationship drive the violations (p > 0.000).

An auto-fitted generalized ordered logit model was used to allow the effect of severity and relationship status in Model D of Table 4 to vary by outcome category (Long & Freese, 2006). As Agga and Scott (2015, p. 376) note, the generalized ordered logit model is equivalent to a “series of binary logistic regressions where the categories of the dependent variable are dichotomized at each cut off.” This model does not incorporate stringent proportionality assumptions. Regarding interpretation, positive coefficients indicate higher values of a variable predicted higher values of the outcome. Results of the generalized ordered logit model (see Appendix Table 7) suggest that incident severity is a significant predictor across each outcome comparison; however, the effect is weakened at greater levels of punishment agreement. A similar pattern was found for the relationship status estimates, as respondents in a committed relationship display greater levels of agreement, with less differentiation in coefficients at higher levels of agreement. The effects of the other independent variables of interest remain consistent with the results already presented in the ordered logit analysis.

Table 5 displays the results of analysis separated by the high and low severity conditions for the police notification and jail as punishment outcomes with the comparison to heterosexual couples where women are the victims. When the incident of violence is less severe (Model A), third parties are most likely to notify the police on behalf of female victims in heterosexual relationships, holding all else constant (see Table 5). Furthermore, third parties are also least likely to call the police on behalf of men victimized by their female partners (β − 2.23; p < 0.01). Model B indicates that third parties are most willing to mobilize the law in severe cases when the victim is female and heterosexual, followed by females in same-sex partnerships. Severity does not fundamentally change the nature of the association between couple type and police notification.

Table 5 Ordinal logistic regression predicting notifying police and jail as punishment by incident severity^
Full size table

Figure 1 illustrates the predicted probabilities of police notification for less (panel A) and more severe (panel B) incidents of partner violence. Overall, the willingness of third parties to notify police is greater in severe incidents, regardless of disputant sexuality and gender. For instance, the probability of “likely” notifying police is roughly 0.24 in less severe incidents of a man hitting a woman. The predicted value exceeds 0.60 in more severe incidents. Also, the probability of “unlikely” reporting is lower than “likely” reporting in the severe incidents save for a single couple type: when a man is attacked by a women partner. Here, third parties are still more “unlikely” to report to the police than to report.

Fig. 1
figure 1

Predicted probabilities of notifying the police in less and more severe incidents of violence

Full size image

Figure 2 displays the predicted values for less and more severe incidents when jail as punishment is the outcome variable. Generally, respondents more strongly favor punishment when violent incidents are more severe as opposed to less severe. The degree of preferred criminal punishment, again, seems to be more affected by the gender of the disputants than their sexuality. Support for jail as punishment is higher for female victims in heterosexual relationships who are attacked by a male partner regardless of severity but especially in severe cases. Also, women attacked by women in homosexual relationships receive the next highest level of support, which is stronger in severe incidents. Severity increases punishment preferences for male victims but does little to approach the degree of preferences favoring cases involving female victims.

Fig. 2
figure 2

Predicted probabilities of jail as punishment in less and more severe incidents of partner violence

Full size image

As a specification check, we display the marginal estimated effects for each condition across the moderator variable of incident severity, shown in Table 6. The results indicate that when the severity of violence is high, third parties are significantly more supportive of notifying police and assigning jail as punishment across all gender comparisons. However, as the core results have demonstrated, third-party bystanders are most likely to mobilize the law on behalf of heterosexual female victims.

Table 6 Marginal effects (predicted probability) for notifying police and jail as punishment by incident severity and couple composition (interaction)
Full size table

As a supplement to the primary analysis, it is possible that punishment preferences might account for the effects of gender but also have direct associations with police notification. For exploratory purposes only, the jail as punishment outcome measure was included as a mediator in the full ordinal logit model predicting police notification. This analysis revealed that when respondents believe an aggressor deserves jail as punishment, they are more willing to notify the police of the incident (β 0.98; p < 0.01). This finding implies that a third party’s decision to mobilize the law is partly affected by how deserving of punishment they find a disputant. Incidentally, these results suggest a strong degree of convergence between the two legal mobilization preferences (see Kidd, 1979)—between the punishment and notification outcomes. Notably, the direction and strength of the coefficients for the other variables are very similar with the inclusion of the punishment measure, and thus, the substantive conclusions remain.

Discussion

The current study sought to determine the extent to which the sexuality and gender of intimate partners in a violent dispute affect whether third parties notify the police and prefer criminal punishment. Furthermore, the study assessed how the potential effects of these conditions depend on the severity of violence between intimate partners. The analyses uncovered several findings from an experimental vignette design which have implications for both criminological research and crime control policy.

  1. o

    First, the sexuality of disputants is not the primary factor affecting the decisions by third parties to involve the criminal justice system. Specifically, initial comparisons between sexual minority and heterosexual couples revealed that third parties were less inclined to notify the police and seek jail as punishment for gay (versus straight) couples. Because the comparison did not partition the gender of the disputants in the heterosexual category, it was not possible to isolate the gender effect from the sexuality effect. As such, respondents may have been comparing incidents involving gay male victims to incidents involving heterosexual females. This initial comparison demonstrates the necessity of accounting for gender to isolate differences in third-party responses to heterosexual and homosexual IPV.

  2. o

    Second, subsequent analyses indicated that the gender of the disputants, and not their sexuality, was the most influential determinant of whether third parties would notify the police and select jail as a punishment. The findings therefore do not support the assertion nor empirical evidence suggesting that IPV in same-sex partnerships elicits fundamentally different responses from third parties than in heterosexual partnerships (e.g., Brown, 2008; Stanziani et al., 2018, 2020). The findings are largely consistent with prior research which has revealed a primary role of gender as opposed to sexuality (Seelau et al., 2003; Sorenson & Thomas, 2009). Also, they do not support our interpretations of Black’s (1978) model suggesting that homosexual couples receive less law than heterosexuals owing to discrepancies in their cultural status.

Respondents were most likely to mobilize the law on behalf of heterosexual women, followed by homosexual women. Lesbian victims elicited less of a third-party response than heterosexual women victims probably because their assailants were women (rather than men). These findings are consistent with claims in the literature suggesting that the public is most responsive to traditional “heteronormative” cases of IPV where a man attacks a woman (Seelau et al., 2003; Seelau & Seelau, 2005; Sorenson & Thomas, 2009) and perhaps because they view these acts as more dangerous (Stanziani et al., 2018). Physical power imbalances may hold important clues about the source of this effect, as men are widely viewed as more threatening and dangerous than women (Felson & Cares, 2005; Hamby & Jackson, 2010; Russell et al., 2015, p. 52). The findings, however, also suggest that a societal norm protecting women from harm (Brown, 2008; Felson, 2000)—a chivalry norm, may extend to lesbian IPV victims given that they were the second most likely to elicit third-party legal mobilization.

  1. o

    Third, the results revealed that third parties were the least likely to mobilize the law in incidents involving heterosexual male victims. Moreover, gay male victims elicited a stronger level of support from third parties than heterosexual males, which is inconsistent with arguments that this group is judged to be less worthy of protection (e.g., Brown, 2008; Russell et al., 2015). Gay men, lesbian women, and heterosexual women attracted stronger third-party support than heterosexual male victims. These findings are consistent with prior research and align with the perspective that victims of IPV “who may be perceived as having less structural power are viewed as being more in need of societal assistance” (Sorenson & Thomas, 2009, p. 348).

  2. o

    Fourth, the results suggest that variation in the incident severity does not diminish the primary effects of gender on either dimension of third-party legal mobilization, nor does severity yield substantively different effects by sexuality. The severity condition essentially amplified but did not fundamentally change existing tendencies to mobilize the law based on the characteristics of the dyads. Namely, severe violence maximizes the effects of gender on third-party reactions. This finding suggests that the costs of non-intervention in severe cases of IPV do not offset nor dilute preexisting normative priorities assigned to gender but instead only amplify them. This pattern might inform an elaboration of Kidd’s (1979) social psychological model. Also, it is worth noting that incident severity had a main effect on police notification and punishment, suggesting that respondents are more inclined to protect victims and punish perpetrators in incidents involving greater (vs. minor) physical harm.

Despite its contributions to the literature, this study is not without some limitations. The utilization of a between-subjects design renders comparisons across conditions by a single respondent impossible. Rather, analyses can only compare subjects that have been randomly assigned to the same conditions. Second, the sample utilized in this study was drawn from Amazon Mechanical Turk (MTurk), thus rendering it a convenience sample. A possible limitation associated with using MTurk sampling procedures include potential differences between workers and the general population on both measured and unmeasured characteristics, thus possibly affecting study’s generalizability from the laboratory to the wider adult population (Goodman et al., 2013). However, as experimental social psychologists have argued, “laboratory experiments do not generalize to the larger population. Rather, they tell us something about the conditions under which they theoretical operate and produce the predicted outcomes” (Horne & Lovaglia, 2008, p. 4). Furthermore, across the models, female respondents reported a greater likelihood of mobilizing the law when witnessing incidents of IPV. As a result, the results of the study might be biased towards responses typical of women respondents given the slight overrepresentation of women in the MTurk sample. Additionally, the sample was more educated than the general population, and more educated individuals are less punitive towards violent criminal offenders, and therefore, the sample may understate the extent to which third parties prefer jail as a punishment for IPV offenders. Also, the possibility remains that the different forms of violence depicted in the vignettes may not be qualitatively dissimilar to produce different bystander intentions across the severity condition. To some respondents, a slap across the face compared to punch and shove to the ground may be equivalent types of physical harm. Or perhaps the violence described here simply does not cross a threshold of seriousness that triggers fundamental changes in how third parties perceive the costs of not intervening on behalf of victims (see Kidd, 1979). This level might require life-threatening injuries to the victim including the use of deadly weapons (e.g., firearms).

Future studies should consider including race as an addition to the vignette conditions (see Sola & Kubrin, 2023). Studies that have examined race and helping behaviors have found that third parties are more willing to help members of their same racial group (Piliavin et al., 1969). Due to a long history of the criminalization of African Americans, respondents may feel that Black victims are less deserving of intervention. A related body of criminal sentencing research has found that homicides involving White victims are more likely to result in a death sentence than homicides with Black victims (Radelet & Pierce, 1991; Sorensen & Wallace, 1995). This pattern indicates greater condemnation of violence against Whites by minorities, and thus, it could be hypothesized that greater support would be found in situations of IPV involving a White victim. Furthermore, incorporating the age of the respondents would permit assessments of whether third-party responses to the sexuality and gender of disputants are age-dependent, perhaps reflecting generational differences in beliefs about the qualities of victim worthiness.

Future research should assess how perceptions of danger vary with the gender and sexuality of the dyad which, in turn, influence third-party legal mobilization intentions. A vignette study focused only on opposite-sex couples found that violence against women is viewed as most threatening because men are seen as more physically powerful than women (Hamby & Jackson, 2010). This same approach, but with comparisons between opposite and same-sex IPV incidents, would be an important contribution. Furthermore, studies should gather data on whether respondents believe in a norm protecting women from harm (see Felson, 2002) and test if this norm affects their responses to IPV incidents. As noted, gender effects have been well-documented in research on third-party reactions to IPV among opposite-sex couples and to a small extent in same-sex couples. Yet there has been very limited research on the social psychological and cognitive elements that account for these effects. Lastly, the vignette developed for the current study held constant conditions for the relationship (married) between disputants, the setting (public), as well as closeness of the respondent to the parties in conflict (acquaintance), thus was unable to detect effects of variation in the location, relationship type, and relational distance (see West et al., 2023). Allowing for variability in these dimensions would permit a more faithful test of Black’s (1978) model than the present design permits. However, future studies should consider varying the setting of the IPV incident and the relationship of the disputants (to each other and the third party), while including conditions for sexuality.

The findings have implications for IPV prevention programs and policies regarding arrest and punishment. What is necessary for effective prevention, arrest, and sanctioning is the consent from the public that IPV is not only harmful but warrants criminal justice intervention. There is, no doubt, a strong consensus among the public to support efforts to reduce the burden of IPV. But, as our findings show, this support mainly aligns with heteronormative views of IPV. As noted, bystanders are the least responsive to the plight of male victims of IPV. By extension, this pattern suggests that male victims of IPV are less likely to receive victim services, to receive support from the community, and be protected by the law from subsequent victimization (Briones-Robinson et al., 2016). Their assailants, both women and men, can essentially attack them with impunity. Public health campaigns have done a great deal to raise awareness of the harms of partner violence, to encourage bystander intervention (Estefan et al., 2019; Spivak et al., 2014). Similar attention should be increasingly directed at raising public awareness about the scope of violence against men in heterosexual and homosexual relationships—known in public health parlance as “changing knowledge and attitudes” (Spivak et al., 2014). Such efforts should aim to counteract stereotypes—rooted in societal masculinity norms—of male invulnerability and toughness in order to foster societal recognition of men as victims. At the same time, law enforcement officials and advocacy organizations must develop policies upholding their responsivity to male IPV victims as way to foster equal treatment regardless of gender.

Conclusion

Given that victims of intimate partner violence are reluctant to report to the police, examining factors that influence the third-party response to situations of IPV is important. The present study addressed key gaps in the literature, indicating that gender and severity, rather than sexuality, are the most salient predictors of a third party’s intention to mobilize the law, with greater intention to aid female victims. With the growing interest in criminology of those within the LGBTQ + community, this study provides insight into the response of third parties when confronting IPV among same-sex couples. Results of this study warrant further analysis of queer relationships and violence, as well as how they are perceived by the public. The response of third parties, both informal and formal, is a crucial factor in the study of partner violence more broadly, and this literature can be expanded to further understand queer relationships.