Does sexual orientation of bullying target influence bystander response?

There is a dearth of research in the organizational sciences relative to how and why a bystander to workplace bullying may respond based on the sexual orientation of the target. The primary aim of the current study is to explore this line of questioning with a specific focus on women targets.

Drawing from research in education and social psychology, we use the Bystander Intervention Model (BIM; Latané and Darley, 1970), which has not been tested in working adults (except in the context of cyberbullying in college students) (Brody and Vangelisti, 2016), as we wish to discover its applicability to workplace bullying. Further, we extend the BIM by including covariates of homophobia and amnestic heterosexism, discussed in the following sections. The proposed relationships are illustrated in Figure 1.

Figure 1

We agree with Worthen’s (2013) compelling argument that research should distinguish among sexual minority groups, since lesbians and gay men experience workplace bullying differently (Salin and Hoel, 2013). To date, much of the research on sexual minority populations has failed to differentiate among these subgroups (Herek, 2002; Worthen, 2013), which risks overlooking important differences within these populations (Herek, 1994; Katz-Wise and Hyde, 2012; MacInnis and Hodson, 2015). We address this deficiency by investigating three groups of women (lesbians, bisexuals, and heterosexuals) and thus avoid aggregating sexual minority individuals into a single category dubbed “LGB”.

Female targets. We study female targets because they are often disadvantaged in many settings. Since most workplaces continue to be masculinized environments (Berdahl, et al., 2018), women occupy lower-impact jobs (DeSilver, 2018) and receive lower pay (Graf, et al., 2018), making them more vulnerable to mistreatment (Gates and Viggiani, 2013). Further, sexual minority women in particular may face situations of bullying at a higher rate than heterosexual women. Thus, our study of women who are targeted by bullies is framed by the theories of homophobia (fear of homosexuality; Fyfe, 1983; Herek, 1995) and amnestic heterosexism (belief that sexual minorities are not disadvantaged; Walls, 2008) as they help to explain why sexual orientation of the target may be a salient cue for bystander responses. We extend Katz et al.’s (2019) study of anti-gay (male) bullying to a different target population, women targets – lesbians, bisexuals, and heterosexuals.

Bisexual targets. Research on bisexuals is especially sparse (Corrington, et al., 2018) despite the fact that they constitute the largest reported segment of the LGBT population (Catalyst, 2017; Gallup, 2017). Perceptions of bisexual individuals are distinct from those of homosexual individuals and it has been suggested that attitudes towards each subgroup are unique constructs (Eliason, 1997). Thus, we explore the perception of bystanders regarding bullying of bisexual (females) as a specific subsample within the study. We offer research questions as there is insufficient previous research to allow us to formulate testable hypotheses about this segment of the sexual minority population.

Bystanders to emergencies, crises, and abuse may respond in ways that escalate or halt the situation. While bystander responses to crises and emergencies have been studied in violence/sexual assault (Chabot et al., 2018; Lukacena et al., 2019), medical emergency (Fischer et al., 2011), cyberbullying (Koehler and Weber, 2018), and school bullying settings (Jenkins and Nickerson, 2019), there is far less in the workplace bullying context (Mulder et al., 2016; Paull et al., 2012), especially related to target sexual orientation as an antecedant.

Noteworthy research on bystanders includes Hellemans et al.’s (2017) study that found the most common bystander responses to bullying were either passive or noninterventionist. MacCurtain et al.’s (2018) study of Irish nurses who witnessed bullying found that 64% adopted a low-involvement intervention (i.e. informally discuss the situation with colleagues or superior) while 9% did nothing in response. Bullying mode (online or offline), bullying type (personal or work-focused), and target-bystander relationship influenced bystander intentions to intervene in Coyne et al.’s (2019) cross-sectional study. Lastly, Mulder et al. (2014) found that bystander intervention was, in part, determined by ascribed responsibility for the bullying behavior. Other studies on bystanders have focused on the impact to themselves as witnesses. Negative consequences such as depression (Emdad et al., 2013) and stress (Vartia, 2001) have been reported.

The BIM (Latané and Darley, 1970) has been utilized in studies of sexual harassment (Nickerson et al., 2014), domestic violence (Hoefnagels and Zwikker, 2001), and emergencies (Darley et al, 1973). The BIM (Latané and Darley, 1970) progresses through five sequential stages that we situate within a workplace bullying context. First, the bystander must notice the event; thus, the process begins with awareness that bullying is occurring. Next, the bystander must interpret the event as an emergency requiring help; the bullying must be assessed and judged to be sufficiently harmful to the target. Third, the bystander accepts responsibility for intervening. No longer considered a disinterested third party, the bystander determines that s/he has a role to play in addressing the bullying. Next, the bystander must know what to do and how to intercede. This presumes that the bystander knows how to respond to apparent bullying. Finally, the bystander implements an intervention decision, choosing from various options.

Myriad factors have been shown to influence the bystander’s propensity to act – gender (Poteat and Vecho, 2016), relationship/friendship with the target (Coyne et al., 2019; Ryan and Wessel, 2012; D’Cruz and Noranda, 2010), perceived severity of the incident (Koehler and Weber, 2018), attributed target responsibility (Mulder et al., 2014), empathy (Nickerson et al., 2014) and accomplices (Namie and Lutgen-Sandvik, 2010). Bystander responses to emergency situations are complex and involve cognitive, emotional, and behavioral components (Koehler and Weber, 2018). Additionally, the context of the situation may also play a part in influencing bystander perceptions and decisions.

A contextual factor that has been shown to influence bystander responses to an emergency is the ambiguity or clarity of the situation (Nickerson et al., 2014). Extant research by Latane and Darley (1968) showed that the interpretation of an ambiguous situation (e.g. a smoke-filled room) to be an emergency was influenced by other witnesses to the event. In their study of alcohol-related sexual assault on college campuses, Pugh et al. (2016) found that the certainty of risk to a female victim was one element in determining bystanders’ response to the situation. When bystanders determined that the situation constituted a clear, blatant emergency, they were more likely to intervene than when the victim did not appear to be in immediate danger. Observers of sexual orientation harassment were more likely to intercede when the perpetrator’s intent to harm the target was unequivocal (Ryan and Wessel, 2012). The absence of strong, compelling signals to act may reduce bystander accountability in ambiguous situations (Latané and Darley, 1970). Therefore, the explicitness or ambiguity of an event may affect the bystander’s subsequent decisions and actions.

Homophobia is a fear of homosexuality that often results in negative stereotypes and discrimination against those in the sexual minority (Fyfe, 1983; Herek, 1995). It is frequently associated with derogatory remarks and treatment of LGB individuals (Einarsdóttir et al., 2015; Hettinger and Vandello, 2014). The construct of homophobia is well explored and researched in the sociology and gender studies literatures with less attention focused on it in the organization sciences, particularly studies of the workplace (Eliason et al., 2011).

Homophobia is pervasive and can lead to mistreatment of individuals who are either openly homosexual or are perceived to be homosexual. For example, prejudice and discrimination in employment decisions are not uncommon, with 27% of LGB workers (Katz-Wise and Hyde, 2012) and 38% of openly homosexual workers reporting experiencing at least one form of sexual orientation discrimination (Sears and Mallory, 2011). Studies of selective incivility provide further evidence that sexual minority workers are subjected to higher levels of workplace incivility than their heterosexual counterparts (Zurbrügg and Miner, 2016).

Amnestic heterosexism (AH) is the belief that individuals in the sexual minority are no longer subjected to discrimination related to their sexual orientation (Katz et al., 2019). Derived from the term ‘amnesia,’ individuals who hold AH beliefs may deny that LGB people face discrimination or unfair treatment because of sexual minorities’ perceived progress in advancing their rights (Thorpe, 2017). In other words, these individuals have “forgotten” that LGB individuals were once mistreated and now presume that there is no ongoing issue. This leads them to dismiss the existence of sexual orientation discrimination. On its surface, this may seem counterintuitive, but is understandable based on cognitive dissonance theory (Festinger, 1962; Katz, et al., 2019). Those who believe LGB individuals are now on an “equal playing field” with heterosexuals may experience cognitive dissonance when they witness LGB-focused discrimination. Since they believe this type of discrimination no longer exists, they resolve their cognitive dissonance by removing themselves from the situation – in bystanders’ case, by refusing to intervene or blaming the victim. Thus, witnesses higher in AH will be less likely to see themselves responsible for intervention, or to intervene in the bullying.

Indeed, AH has been shown to mediate the relationship between support for gay rights and ‘belief in a just world’ (i.e. people get what they deserve; Hettinger and Vandello, 2014). Notably, Katz et al. (2019) found that those higher in AH felt lower levels of responsibility to intervene in bullying directed at gay men. AH is therefore a potential explanatory variable in understanding why bystanders to a workplace bullying incident may refrain from interceding, since they would not view the situation as mistreatment or harmful.

We believe that these two attitudes, homophobia and AH, will affect how bystanders respond to witnessing an ambiguous workplace situation involving women of various sexual orientations. We study these dynamics in the context of the BIM (Latané and Darley, 1970). The first stage of the BIM is awareness that bullying is happening. Since this awareness is presented in our experimental manipulation, we begin our first set of hypotheses with the BIM’s second stage, interpreting the behavior as bullying.

Note that we refer to “mistreatment” in our hypotheses, as this is the term we used in research materials seen by participants. We did not use the term “bullying” in research materials, since “bullying” is now a commonly used and emotionally charged word that may trigger certain responses; we desired participants to respond to the behavior without automatically defining it as “bullying” to avoid the possibility of priming specific responses. For clarity, we continued this word choice in the method and analyses sections.

As noted above, the lack of research on bisexuals prevents us from making specific predictions about this subsample. Thus, we test research questions related to each hypothesis.

We now address later stages of the BIM, beginning with accepting responsibility for intervening (stage three). The precursor to taking any action is embracing the responsibility to act. Diffusion of responsibility is an enduring obstacle (Beyer et al., 2017) as the number of other bystanders is a powerful influence on felt responsibility (Cox and Adam, 2018; Katz and Klainberg, 2019). Considerable research has shown that as the number of additional bystanders increases, the felt responsibility for a lone bystander diminishes (the so-called ‘bystander effect’).

Accepting responsibility to intervene may also be influenced by the severity of the situation (Hellemans et al., 2017), the bystander’s gender (e.g. female) (Bennett et al., 2017), and the relationship with the target and/or perpetrator (Palmer et al., 2018). In-group targets of violence were more likely to be helped by bystanders than out-group targets in experimental research conducted in the U.K. (Levine et al., 2002). Considering that sexual minority targets of mistreatment would be more likely considered as out-group individuals, we predict the following:

Hypothesis 2: Bystanders’ attitudes toward sexual minority females will predict their intention to intervene when faced with ambiguous negative behavior. Specifically:

H2a: When faced with ambiguous negative behavior, bystanders high in homophobia will be less likely to feel they have personal responsibility to intervene when the target is lesbian and more likely to feel they have personal responsibility to intervene when the target is heterosexual.

Research question 2: When faced with ambiguous negative behavior, how will bystanders’ attitudes towards bisexuals predict whether they feel personal responsibility to intervene?

Lastly, we address the final (fifth) stage of the BIM, the type of intervention decision. (We omit testing the fourth stage, knowing what to do to intervene; since we limited participants to full-time employed adults 25 or over, we assume they will be familiar with normal workplace interventions). Intervention alternatives include more and less assertive responses. The bystander might choose to directly confront the perpetrator, a more assertive option, or indirectly intercede by reporting the situation to someone in authority, a less assertive response. Limited prior research could be located that specifically focused on the assertiveness of the bystander’s response choice.

Bystander intention to intervene assertively may be influenced by the potential costs incurred as a result. Ryan and Wessel (2012) found that observers of workplace sexual orientation harassment were influenced by the possibility that failure to intervene would lead to a generalized hostile work environment that would also impact their working conditions. As detailed in the arousal-cost-reward model (Piliavin et al., 1975), bystanders weigh the costs of helping against those of not helping and then act such that their own net costs are minimized (Fritzsche et al., 2000). Thus, we predict that the intervention chosen will reflect minimal cost to the bystander.

Hypothesis 3: Bystanders’ attitudes toward sexual minority females will predict their reported level of intended interventions when faced with ambiguous negative behavior. Specifically,

H3a: Bystanders high in homophobia will report an intention to take less active interventions when the target is lesbian and to take more active interventions when the target is heterosexual.

Research question 3: When faced with ambiguous negative behavior, how will bystanders’ attitudes toward bisexuals predict how actively they report they would intervene?

The pre-test was designed to identify vignettes and ensure the effectiveness of the experimental manipulation of the target’s sexual orientation. Vignettes were developed by first referring to Escartin et al.’s (2009) identification of bullying behavior severity. Our goal was to choose vignettes that were ambiguous enough to require respondents to evaluate event severity; extreme examples of bullying would elicit clear, socially expected responses, which would not allow us to test our hypotheses adequately. We chose “manipulating information,” “abusive working conditions,” and “professional discredit,” all of which were ranked as the “middle level” of severity, and developed four vignettes illustrating examples of these behaviors (see Appendix A). We constructed vignettes to match the definition of bullying: they include behaviors that offend, exclude, or negatively affect the target’s work and are repeated and regular (Zapf et al., 2002). The pre-test utilized 50 Mturk respondents (Amazon’s online data collection methodology). The sexual orientation of the target (“Susan”) was manipulated by describing her as dating “Thomas” (heterosexual condition), “Maria” (lesbian condition), or “Maria and Thomas” (bisexual condition) as well as clearly stating Susan’s orientation (the manipulation can be found in Appendix A). The pre-test respondents identified Susan’s correct sexual orientation 81 percent of the time.

The pre-test also identified two vignettes used for the study, based on participants’ responses to two questions: “how clearly does this vignette represent workplace mistreatment” and “how harmful does Terry’s behavior in the vignette seem?” We chose vignettes based on these criteria: 1) vignettes rated as most ambiguous (those whose means were closest to the middle of the four item scale); vignettes whose interpretations were most variable (those whose standard deviations were largest). The vignettes chosen are labeled #1 and #2 in Appendix A.

We randomly presented respondents with one of the three sexual orientations and they read and reacted to the two vignettes. Since responses to both vignettes represented each person’s typical bystander reactions and intentions, we combined each respondent’s data across the two vignettes for all analyses.

Clarity and harmfulness of mistreatment (hypothesis 1) were measured identically to the pre-test (one item each; Katz et al., 2019). We operationalized intention to intervene (hypothesis 2) by using personal responsibility to help, as measured by Fischer et al.’s (2006) four-item scale (sample item: “I feel personally responsible for helping Susan in this situation”; α = .95).

The level of intended intervention (hypothesis 3) was adapted by using nine items from Salmivalli, et al.’s (1996) study of school bullying; Nickerson, et al. (2014) adapted these items in a similar fashion. Three items each represented behaviors 1) defending the target (sample item: “I would tell Terry I’m going to tell George about his behavior toward Susan”), 2) sympathizing with the target (sample item: I would make a point to spend more time with Susan”), and 3) avoiding the problem (sample item: “I would stay out of it”). Because of our sample size, we wanted to reduce scale dimensions where possible, thus we factor analyzed type of intervention (maximum likelihood, varimax rotation), resulting in two clear factors (accounting for 48 percent of the variance) that emerged for each vignette: avoidance (three items; α = .88) and intervention (six items, α = .88).

Homophobia was measured by Raja and Stokes’ (1998) seven-item Homophobia Scale (sample item: “Homosexuals are just as moral as heterosexuals”; α = .94). AH was measured using the four items from the Multidimensional Heterosexism Inventory’s AH subscale (Walls, 2008) (sample item: “Most people treat lesbians as fair as they treat everyone else”; α = .93).

The sample for the study was obtained through Mturk, Amazon’s online data collection methodology. Although online data collection methodologies are growing in acceptance, it remains important to carefully manage this type of data collection to ensure validity of responses. A frequent concern is that “professional responders” will participate in multiple surveys as soon as they are available (Walter et al., 2016). To determine if this dynamic occurred in our data, we divided the adult sample into four equal groups based on order of submission and compared the dependent variables’ means. Finding no pattern of differences between “first responders” and later responders, we eliminated order effect as a threat to validity. We also set the software (Qualtrics^®) to reject multiple responses from the same respondent and ensured respondents’ time to complete the survey was adequate. Following new research indicating that attention checks (designed to ensure participants do not mindlessly check boxes) are detrimental to online studies (Vanette, 2017), we opted instead for asking respondents to agree they would thoughtfully answer each question. In both data collections, only U.S. adults (age 25 or over) who were employed full-time (at least 30 hours a week) were eligible.

After eliminating 35 respondents for missing data or failing the manipulation check, 168 respondents were available for Study 2. Their average age was 33, with a minimum of 25 and maximum of 90. Most were male (62 percent); a very small number (3 respondents) reported they were transgender or “other.” The majority were heterosexual (80 percent), with 16 percent reporting being bisexual, one person reported being a gay man, two respondents being lesbians, and one reported being “queer/questioning.”

Hypothesis 1 predicts that respondents who are higher in homophobia (1a) or AH (1b) will be less likely to interpret a lesbian target’s experience as clearly mistreatment (operationalized as bullying) and less likely to see it harmful than they will the heterosexual target’s experiences. Although as Table 2 (below) indicates no significant differences are seen between the conditions on either dependent variable, including homophobia or AH as covariates (thus controlling for them) does not change these results, as can be seen in Table 3. Regardless of whether the target is lesbian or heterosexual, respondents higher in homophobia do not differ in how clearly the behavior represents mistreatment (ρ = .562) or how harmful it is (ρ = .410). Similarly, respondents higher in AH also do not distinguish between these two variables based on sexual orientation (clarity, ρ =. 433; harmfulness, ρ =. 396). Thus, neither hypothesis 1a nor 1b is supported.

Research question 1 explores if respondents will interpret behavior differently if the target is bisexual as compared to heterosexual. Table 3 also shows these results. Regardless of their levels of homophobia or AH, our respondents do not distinguish between bisexuals and heterosexuals in the clarity or harmfulness of negative behavior.

Hypothesis 2 explores if homophobia (2a) or AH (2b) affect intention to intervene, as measured by feelings of personal responsibility toward the target. Our analysis of main effects (Table 2, above) illustrates that personal responsibility differs across the target’s sexual orientation at a less-than-significant level (ρ = .067). Including homophobia as a covariate, however, shows more homophobic respondents tend to feel less personal responsibility when the target of mistreatment is lesbian (Table 4; ρ = .041). The same result is shown for hypothesis 2b: respondents higher in AH tend feel less personal responsibility when the target of mistreatment is lesbian (Table 4, ρ = .027). Thus, hypothesis 2 is supported.

Research question 2 explores this same dynamic regarding bisexual targets. As can be seen in Table 4, respondents higher in homophobia or AH did not differ in their feelings of personal responsibility based on whether the target was heterosexual or bisexual (ρ =. 930 and ρ =. 801, respectively).

Hypothesis 3 predicts that respondents who are higher in homophobia (3a) and AH (3b) will be less likely to intend to actively intervene in mistreatment when the target is lesbian. As discussed, we measured intervention in two dimensions: avoidance and active intervention intentions. As noted above (Table 2), no differences between the target’s sexual orientations are seen in avoidance (ρ = .423), but overall, respondents report significantly lower intentions for active interventions when the target is lesbian as compared to heterosexual (ρ = .031). Investigating this relationship in the context of our covariates, we find that respondents higher in homophobia do not differ in intentions to avoid intervention for lesbian targets (ρ = .728), but they do report lower intentions to actively intervene when targets are lesbian (ρ = .015).

Repeating these analyses with AH as the covariate similarly does not reveal higher avoidance intention for lesbian targets (Table 5, ρ = .999), but does show a tendency for less active intervention for lesbian as compared to heterosexual targets (ρ = .009). Thus, hypothesis 3 is partially supported.

The third research question investigates these relationships when targets are bisexual. Our results show that respondents higher in either homophobia or AH do not distinguish between heterosexuals and bisexuals in terms of their invention intentions, either to avoid or to intervene (ρ =. 429 and ρ =. 368, respectively).

In reviewing our data post-hoc, an analysis of variance of homophobia and AH across sexual orientation conditions revealed significant differences – surprising given that random assignment should equalize all personal characteristics across conditions. The only reasonable conclusion is that mere assignment to the sexual minority condition created a social desirability bias; knowing the target of bullying was a sexual minority lowered responses to the homophobia and AH scales. Compared to respondents whose target was heterosexual, respondents whose target was lesbian were significantly lower in homophobia (ρ < .039) and AH (ρ < .008) and those whose target was bisexual was significantly lower in AH (ρ < .050). On reflection, this is not unexpected, since contemporary U.S. employers emphasize acceptance of diverse sexual orientations. Thus, knowing the target is lesbian or bisexual seems to encourage the desire to “look good”; given these results, it is surprising that some of our hypotheses were supported, as the effects of homophobia and AH seem to have been artificially attenuated.

These findings are of value to organizational leaders for a number of reasons. The literature clearly demonstrates the challenging experiences of those in the sexual minority with respect to workplace treatment. Our research demonstrates the organizational “roadblocks” which might be in place and offers leaders a lens into where they might focus their anti-bullying training protocols. For example, as respondents did not differ in their interpretation of a bullying event based on the sexual orientation of the target, negative behaviors seem to be interpreted similarly, regardless of the target. However, when it came to their feelings of personal responsibility to act on behalf of the target, respondents higher in homophobia or AH felt less responsible when the target was lesbian. This is important for organizational leaders to understand, recognize, and address before bullying happens within their company. It suggests that heightening respect and understanding of sexual minorities is important in motivating bystanders to help targets of bullying; interventions such as affinity group interactions, “safe space” training, or other diversity education may increase feelings of personal responsibility for action against bullying regardless of sexual orientation.

Finally, organizational leaders should recognize that bystanders of bullying may actively intervene less often when the target is a sexual minority, particularly when bystanders hold homophobic or amnestic heterosexist attitudes. Leaders should consider ways to mitigate instances where a sexual minority employee might be bullied at work and determine how to encourage bystanders to intercede.

Contributions, limitations, and future research

We believe this study contributes to the literature in several ways. First, it applies the BIM, a model frequently found in studies of non-work-based settings, to the workplace. This has been done only rarely in past research. We also show that the bystander experience is appropriately framed by this model.

Second, we separated sexual minority subgroups (lesbians and female bisexuals), consequently finding different results. This is an advance in the very small body of literature that addresses bisexuals as a group different from other sexual minorities and worthy of study.

Third, we show that lesbian targets of bullying are indeed viewed differently by bystanders, and that bystanders themselves differ in their propensity to intervene. We found that bystanders who are more homophobic and amnestic heterosexist are less likely to take personal responsibility or actively intervene when they witness lesbian bullying targets. Thus, we show that these variables are important attitudes that are predictive of bystander responses to the bullying of sexual minorities.

The contributions of this research should be considered in light of the limitations. For example, though we originally sought to collect data which would offer a sufficient sample size, this did not prove to be the case. As can be seen, the partial η²’s for many of the equations were low, ranging from below .01 (no effect) to .04 (a medium effect). Unfortunately, resources did not allow for a larger sample size, which we believe would have bolstered our results. Given that even with low power we saw significant and meaningful findings, larger sample sizes are warranted.

A further weakness is that our study presents “paper people” to respondents. They did not view real-life interactions. A laboratory study using videos of vignettes would be a more externally valid research design.

Future research should, along with an increase in sample size, apply a more robust research design to mitigate potential issues with social desirability bias. While we designed the questionnaire so that respondents were exposed to the vignettes first, and then the attitudes (so that attitudes would not influence their bystander responses), clearly this is a less than optimal design. A better design would be to create a time gap between the measure of bystander reactions and attitudes, which would necessitate surveying the same respondents twice, in two different waves. Our past experience, and that of most researchers, is that this requires significant investment in sample size (due to Wave 2 dropouts). Unfortunately, given the time and budget constraints of our study this was not possible.