Microaggressions are subtle, yet harmful, unintentional or sometimes intentional, verbal and non-verbal messages that individuals from marginalized groups receive due to their marginalized identity/identities (e.g., gender identity, sexual orientation, race, size, etc.; Nadal et al., 2014). In the US, racial microaggressions are frequent and harmful verbal, behavioral, and environmental messages that are experienced by People of Color (Nadal et al., 2014; Sue et al., 2007). Research has demonstrated that racial microaggressions can be detrimental to students’ mental health (Nadal et al., 2014; Sue et al., 2007), as well as their academic and career outcomes (Compton-Lilly, 2020; Keels et al., 2017). Racial microaggressions should—and can—be addressed by bystanders (i.e., third-party individuals) witnessing these events. However, little empirical research has focused on ways to prevent these incidents or how to intervene when they occur. Rather, prior studies have reported on the incidence and effects of racial microaggressions among Black (e.g., Sue et al., 2007), Asian (e.g., Huynh, 2012), Latina/o/x (Minikel-Lacocque, 2013), and Native American (e.g., Johnston-Goodstar & VeLure Roholt, 2017) students. Despite abundant evidence that racial microaggressions are frequent and a harmful force for inequity in higher education, there are key gaps in our understanding of how racial microaggressions can be disrupted by bystanders witnessing racial microaggressions.

College Student Mental Health

College students’ mental health has received increased attention in the 2010s (Eisenberg, 2019), with more presence of more acute health challenges and unmet needs shown among students of color (SOC; Lipson et al., 2018). First-generation college students, who are disproportionately SOC, have been found to have higher rates of stress and depression as compared to their more socioeconomically advantaged peers, but use campus counseling services less often (Stebleton, et al., 2014). These challenges appear to have been exacerbated by the compounding crises that began in the spring and summer of 2020, including but not limited to the COVID-19 pandemic and the collective trauma following the murders of George Floyd, Breonna Taylor, and many other individuals of color in the US (Jones et al., 2022; Soria & Horgos, 2021; Yeh et al., 2022). Crucially, these events have made campus climates stressful environments for SOC who have to navigate both systemic injustices embedded in the structures of higher education (Cabrera, 2019; Sills et al., 2020) and interpersonal forms of aggression from their peers, faculty and staff, and other members of their campus community. These incidents frequently occur as racial microaggressions for African American college students (Solorzano et al., 2000), Indigenous college students (Tachine et al., 2017), and Latina/o/x college students (Yosso et al., 2009). To increase the representation of SOC on college campuses finding ways to address racial microaggressions is critical.

While the term racial “microaggressions” may suggest these acts are insignificant to study, decades of research finds that indeed these brief communications which SOC encounter regularly – even daily – (see Sue et al., 2007; Pierce, 1970) can have prolonged negative effects on mental health (Hernández & Villodas, 2020; Nadal et al., 2014). Moreover, racial microaggressions have been linked in the literature to poor health outcomes for SOC. For example, Black college students are negatively affected by racial microaggressions and may engage in behaviors independently associated with negative mental health, such as risky sexual behavior and excessive drinking and alcohol problems (Blume et al., 2012; Marks & Çiftçi, 2019; Marks et al., 2021). Given these patterns, it is important to consider the potential for interventions that could respond to and intervene in the reproduction of behaviors that perpetuates a racist culture in higher educational spaces, even if the perpetrators are not consciously intending to do harm. Theorists have postulated that racial microaggressions may be more harmful than overt forms of racial discrimination because friends, families, and people not meaning harm may be perpetrators (Sue et al., 2007). Negative messages received from trusted others are internalized differently that those received from people who are not well-intentioned (Sue et al., 2007). Furthermore, the subtle nature of most racial microaggressions may lead the target to question whether they are over-reacting or being overly sensitive, which inevitably involves harmful rumination over the experience (Sue et al., 2007). However, only some preliminary research has found differences, and these differences varied based on the target’s racial identity. For example, Lui (2020) found that the correlation between racial microaggressions and psychological distress was stronger for Asian American and Black American participants than the correlation between overt racial discrimination and psychological distress. However, this finding was not true for Latina/o/x Americans and whether these differences were statistically significant was not examined.

Conceptual Framework: Microintervention Theory and the Bystander Intervention Model

This study combines Sue et al.’s microintervention strategies (Sue et al., 2019) and the bystander intervention model (Latané & Darley, 1970; Nickerson et al., 2014) in its conceptual framework. Microinterventions are defined as positive daily verbal or non-verbal actions that convey to targets of microaggressions: (a) validation of their experiential reality, (b) value as a person, (c) affirmation of their racial or group identity, (d) support and encouragement, and (e) reassurance that they are not alone (Sue et al., 2019). Microinterventions offer a way for bystanders to engage in support and allyship when they witness a microaggression regardless of their racial background. Allies are individuals who possess the power to question, challenge, and even redefine the norms that perpetuate oppressive symptoms (Sue et al., 2019). Bystanders can engage in allyship by preventing racial microaggressions from occurring and/or providing support to a target. To date, this mostly theoretical area of research presents the potential to reduce racial microaggressions in learning environments for SOC by engaging bystanders.

According to Latané and Darley (1970), the bystander effect states that a person’s likelihood of helping someone else decreases when other bystanders are present. These researchers sought to explain why bystanders might help in emergencies. Therefore, they proposed a theoretical model to study psychological processes that may prevent a bystander from helping in an “emergency.” Their studies of bystanders in emergency situations led to the development of a five-step model that ends with intervention in “emergencies,” which has been extended to any problematic interaction (e.g., bullying, harassment). Latané and Darley (1970) posited that intervention by bystanders progresses sequentially in the five-step model. The model includes: (1) noticing the event (i.e., Notice), (2) interpreting the event as one requiring intervention (i.e., Interpret), (3) accepting responsibility for intervening in the event (i.e., Accept), (4) knowing how to intervene (i.e., Know), and (5) implementing decisions regarding intervening (i.e., Act). In sum, the model presents a way in which to conceptualize bystanders intervention when they witness a problematic interaction between two or more people.

Despite the importance of bystander intervention in reducing discriminatory incidents, there are a myriad of reasons why bystanders choose not to intervene. Depending on the number of bystanders there may be a diffusion of responsibility where none of the bystanders intervene because they believe one of the other witnesses will (Fischer et al., 2011). Bystanders may fear the consequences of intervening or be unsure of whether the situation warrants intervention (Murrell, 2021). For bystanders of color, there may be additional deterrents if the perpetrator is White. For example, perhaps they may fear that intervening may make the situation worse for the target or bring them into a harmful situation, engaging aspects of the cost-reward model (Dovidio et al., 2016). Perhaps, intervening as a person of color may make the target more defensive than if the intervener were a White individual.

Hyers (2007) explored bystander intervention experiences in 98 female participants from a Northeastern University and found that although 75% participants wanted to assertively respond in discriminatory experiences, only 40% actually did so. A recent study by Hurd et al. (2022) found that White peers were more likely to intervene in an online racial discrimination incident if they have awareness about the harm it caused to their Black peers, in addition to having the correct perception about social norms that aids confrontation of racial discrimination as well as had the guidance about what to say in these situations. In general, research has repeatedly suggested that the five-step model is a useful framework for understanding bystander behavior. However, this framework has not been adequately studied in individuals who are bystanders in racial microaggression situations. Further, intervening in situations that are microaggressive in nature is different from knowing what can be done in events that involve overt racial discrimination. Noticing microaggressions may be difficult because of their subtle and implicit nature that might be perpetrated unknowingly, such as in the form of a compliment (Sue et al., 2007). Interpreting a racial microaggression as an event that requires active intervention hence could be challenging without having an understanding about the nature of microaggressions, the impact they hold, or even diffusion of responsibility where it is assumed that the situation or event will be intervened by someone else also witnessing the same event (Latané & Darley, 1970).

Predictors of Bystander Intervention (Empathy, Moral Disengagement, and Efficacy)

There is limited research applying the bystander intervention framework to understanding racial microaggressions, but this framework has been applied to understand bystander intervention in bullying (e.g., Jenkins et al., 2018), which is a related social aggression. Although racial microaggressions are different to incidences of bullying, there is some overlap. Both behaviors are aggressive and negatively impact a target who often holds more power in an interaction than the perpetrator. Often both types of incidents occur in situations where there is a bystander, even when the activities occur online (and both can). Allison and Bussey (2016) conducted a review of the literature on bystander intervention in cyberbullying and identified two primary theoretical frameworks that apply to all bystander intervention research. The first framework was the five-step bystander intervention model (described above). The second framework is the social cognitive model (Bandura, 1986), which espouses that personal, behavioral, and environmental factors interact with each other and influence the development of moral standards and moral behavior. In the case of bystander intervention, the moral behavior would be intervening when witnessing an emergency or aggressive act. Allison and Bussey (2016) noted “The apparent failure of moral standards to motivate moral behavior may be mediated by the individual’s cognitions and beliefs, specifically their use of moral disengagement mechanisms and perceived self-efficacy” (p. 188). Allison and Bussey (2016) summarized the literature and noted that moral disengagement, empathy, and self-efficacy are identified as common social cognitive factors that may explain why moral behavior (i.e., bystander intervention) occurs or not, and we believe this model can be applied to bystander intervention in racial microaggressions. The current study thus also uses social cognitive theory to explore the association between moral disengagement, empathy, and self-efficacy to intervene and the steps of the bystander intervention model in the context of racial microaggressions. None of these variables have been examined as predictors of intervention in racial microaggressions. Theory reviewed above support this association, however, as does empirical evidence from the related bullying literature discussed below.

Moral Disengagement

Moral disengagement may sway people from intervening in racial microaggressions, despite their belief that racial microaggressions are hurtful or wrong (Allison & Bussey, 2016). Moral disengagement refers to the process of behaving in a manner not aligned with a person’s moral standards. In the context of racial microaggressions, an individual may recognize the interaction between a perpetrator and a target is wrong but choose not to intervene. Bandura posited eight moral disengagement mechanisms, which fall into four clusters and can be applied to students in college settings. Through cognitive restructuring, students may frame their actions as serving a higher purpose (i.e., moral justification), contrast them to even worst behaviors (i.e., advantageous comparison) or explain them in simple ways (i.e., euphemistic language). They may minimize their responsibility if they were encouraged by others (i.e., displacement of responsibility) or included in a group (i.e., diffusion of responsibility). Students may also minimize the effects of their behaviors by rejecting or understating the effect on targets (i.e., distortion of consequences). Lastly, students may shift focus to targets by demanding provocation (i.e., attribution of blame) or repudiating targets’ humanity (i.e., dehumanization). Research has shown that apathy and moral disengagement are positively associated with passive bystander behavior (i.e., ignoring or pretending not to see bullying; Pozzoli & Gini, 2013) and negatively related to defending (DeSmet et al., 2016). Hence, is seems likely that moral disengagement would be negatively related to each step of the bystander intervention in racial microaggressions model.

Empathy

Empathy refers to the ability to take the perspective of (i.e., cognitive empathy) or emotionally experience (i.e., affective empathy) an event with another person (Allison & Bussey, 2016). In the bullying literature, empathy is a consistent predictor of both traditional and cyber bystander intervention (Fredrick et al., 2020; Ma et al., 2019; Menolascino & Jenkins, 2018), particularly affective empathy. Again and again, intervening in cyberbullying is associated with greater empathy in numerous studies (e.g., Erreygers et al., 2016; Macháčková et al., 2013; Macháčková & Pfetsch, 2016; Van Cleemput et al., 2014). In relation to the five bystander intervention steps, greater affective empathy is associated with interpreting bullying as an emergency and accepting responsibility for intervening. Menolascino and Jenkins (2018) found that higher affective empathy was associated with a greater likelihood of interpreting bullying as an emergency and accepting responsibility for intervening. Hence, is seems likely that this finding may translate to situations in which bystanders witness a racial microaggressive interaction where empathy would be positively related to each step of the bystander intervention in racial microaggressions model.

Self-Efficacy to Intervene

Self-efficacy refers to the belief that one has the confidence and ability to perform a task (Bandura, 1986). Thornberg and Jungert (2013) found that individuals who actively intervene or ignore bullying received similar scores on measures of morality but suggests that self-efficacy is key to intervening. In other words the more confidence around how to intervene a bystander has, the more likely they are to intervene. Many studies have shown that self-efficacy is associated with intervention in bullying (DeSmet et al., 2016; Gahagan et al., 2016; Macháčková et al., 2013; Price et al., 2014), but it is unknown how it is related to each step of the bystander intervention model when applied to racial microaggressions. However, since self-efficacy for intervening increases the likelihood that one will intervene when they witness a problematic event, it seems likely that self-efficacy would be positively related to each step of the bystander intervention in racial microaggressions model.

The Current Study

Racial microaggressions have documented negative effects on the psychological functioning of SOC (Sue et al., 2007) and microinterventions have been theoretically posited as a method of intervening (Sue et al., 2019). The bystander intervention model (Latané & Darley, 1970) is a novel way in which to examine the steps taken by bystanders when they witness a perpetrator engaging in a racial microaggression. Exploring predictors of these steps is important work to ultimately reduce the incidence of racial microaggressions on college campuses through bystander intervention. Hence, the purpose of our research study was to examine potential social cognitive predictors (i.e., moral disengagement, empathy, and self-efficacy) of each of the five steps of the bystander intervention model in a sample of racially diverse college students. Our hypotheses were:

  • H1: Moral disengagement will be negatively associated with each step of the bystander interventions model for White students and SOC.

  • H2: Empathy will be positively associated with each step of the bystander intervention model for White students and SOC.

  • H3: Self-efficacy will be positively associated with each step of the bystander intervention model for White students and SOC.

Method

Participants

There was a total of 452 college student participants ages 19 to 25 years (M = 20.06, SD = 1.18). In the SOC group where were 89 Black students, 88 Latino/a/x students, 38 Asian American students, 29 Multiracial students, five American Indian, Native American, or Alaska Native students, and three Native Hawaiian or Other Pacific Islander students. The remaining 200 participants were White students (i.e., did not identify with any other race and not as Latino/a/x or Hispanic). The sample was 66.6% women, 31.9% men, 1.5% other or they did not identify as either man or woman. Regarding sexual orientation, 86.1% identified as heterosexual, 2.2% gay or lesbian, 7.7% bisexual, 0.4% asexual, 2.7% not sure, and 0.9% preferred not to provide their sexual orientation. Only 1.3% of the sample was married, 27.4% were in a committed long-term partnership, 69.7% were single, and 1.5% said they were in a different situation. Only two participants indicated that they were international students. There were equal numbers of participants who were unemployed (47.3%) and working part-time (47.3%), with 5.3% of the sample working full-time.

Procedures

Prior to data collection, the study and its procedures were approved by the Institutional Review Board at an institution in the Southern United States. To be eligible for the study, participants needed to identify as an undergraduate college student between the ages of 18 and 25 years. Participants were recruited using two mechanisms to increase the racial diversity of the sample. First, data were collected using the undergraduate subject pool in the authors’ college and by sending a mass recruitment email to a random sample of students university-wide at the same institution. Students received course credit for completing the online survey and were entered into a drawing for an Amazon gift card. Second, data were collected using Amazon Mechanical Turk (MTurk). Participants where paid $1.00 for completing the online survey. Consent was gathered electronically prior to the participants beginning the online survey through Qualtrics.

Measures

Bystander Intervention for Racial Microaggressions

The Bystander Intervention Measure for Racial Microaggressions (BIM-RM) was developed for this study by adapting a previously published scale that measured the five-step bystander intervention model in relation to bullying and sexual harassment among high school students (Nickerson et al., 2014). The scale was adapted by changing the wording of the items to reflect bystander intervention in racial microaggressions and providing a definition of racial microaggression. Specifically, the wording of the items was adapted using the microintervention strategies suggested by Sue and colleagues (2019). The adapted scale consists of 26 items across the five subscales: Notice (3 items, e.g., “Racial microaggressions are a problem that I see” and “People I know have experienced racial microaggressions”), Interpret (3 items, e.g., “Racial microaggressions can hurt someone, even if it was unintentional” and “I think racial microaggressions are harmful”), Accept Responsibility (3 items, e.g., “I believe that my actions can help stop racial microaggressions” and “I think it is up to me to help stop racial microaggressions”), Know (3 items, e.g., “I know what to do to get someone to stop engaging in racial microaggressions” and “I have the skills to help someone who is experiencing racial microaggressions”), and Act (14 items, e.g. “If I saw a racial microaggression, I would indicate that the perpetrator said something offensive” and “If I saw a racial microaggression, I would defend the target”). For the Act step, the items were drawn from the list of possible microinterventions suggested by Sue et al. (2019). All items are rated on a four-point scale ranging from 1 = really disagree to 4 = really agree.

A related study using the same dataset (Jenkins et al., 2024) explored psychometric evidence for the new scale. A confirmatory factor analysis revealed good overall fit for the five-factor model (CFI = 0.92, TLI = 0.91, RMSEA = 0.05), and alpha coefficients were moderate but acceptable (0.63 to 0.88). There was also evidence of configural, metric, and scalar measurement invariance when comparing SOC and White participants.

Moral Disengagement

The 8-item Cyberbullying Moral Disengagement Scale (Bussey et al., 2015) was also adapted to racial microaggressions. There is an item for each type of moral disengagement mechanism (i.e., moral justification, euphemistic language, advantageous comparison, displacement of responsibility, diffusion of responsibility, distorting consequences, attribution of blame, dehumanizing), and the questions are specifically tailored to microaggression situations. For example, “It’s alright to say something mean about someone’s race if they have been mean to one of your friends.” and “A racial microaggression is just a way of joking around.” Response options were 1 = totally disagree, 2 = mostly disagree, 3 = mostly agree, and 4 = totally agree. For the original cyberbullying version of the scale, a confirmatory factor analysis indicated that all items loaded onto a single factor and the reliability coefficient was 0.85 (Bussey et al., 2015). In the present study, the reliability coefficient was 0.89.

Empathy

The Interpersonal Reactivity Index (IRI) is a 28-item measure of general empathic tendencies that assesses both cognitive and emotional components of empathy. Items are evaluated on a 5-point Likert scale ranging from 0 = does not describe me well to 4 = describes me very well. This scale yields four subscales, each including seven items, but only the Empathetic Concern subscale was used in the study (i.e., other-oriented feelings of sympathy and concern for unfortunate others; “I am often quite touched by things I see happen” and “When I see someone being treated unfairly, I sometimes don’t feel very much pity for them” which is a reverse-coded item). Items in each subscale are summed to obtain the subscale score (score range for each scale = 0–28), and a high score means higher empathic concern. Internal reliability for the test ranged from 0.71 to 0.77 and test–retest reliability ranged from 0.62 to 0.71. The Empathetic Concern subscale reliability coefficient for the current study fell within that range (0.76).

Self-Efficacy to Intervene

Self-efficacy to intervene was measured by asking participants to rate their ability to engage in each of the intervention options presented in the BIM survey (described above). Bussey et al. (2015) used a similar method to examine self-efficacy to engage in cyberbullying intervention. Participants were given the same list of intervention options (i.e., items 13–26 of the BIM) and asked “For each of these interventions below, how do you rate your ability to engage in each intervention?” For each intervention option, participants rate themselves on a scale of 1 = not at all well to 7 = very well regarding how well they could enact specific behaviors. The original version of the scale had a reliability coefficient of 0.79, but in the current study the reliability coefficient was 0.94.

Data Analyses

IBM SPSS Statistics (version 28) was used to calculate descriptive statistics and intercorrelations among the main study variables. Missing data was less than 5% for all variables. Missing values were coded in the input file and were handled in Mplus by using Full Information Maximum Likelihood. Data screening procedures included examining skewness, kurtosis, and multicollinearity. Collinearity was assessed by examining tolerance and VIF and all values were within the recommended ranges (i.e., tolerance above 0.10 and VIF below 10; Kline, 2011). Kurtosis for Moral Disengagement was elevated (7.39), but was corrected for in the analyses.

To answer the primary research question, path analysis using Mplus 8.0 (Muthén & Muthén, 2017) was used. To account for nonnormality, the robust maximum likelihood (MLR) estimation procedure was utilized. Model fit statistics are not reported because the models were saturated (i.e., all parameters were identified; Little, 2013). Path analysis allows researchers to examine multiple outcomes (i.e., steps of the bystander intervention model) in one model to account for the overlap in these variables. The model had three exogenous variables (i.e., moral disengagement, empathy, and self-efficacy) and five endogenous variables (i.e., Notice, Interpret, Accept, Know, and Act).

A multiple-group path analysis model was tested to see if the model and all path coefficients were the same for White students and SOC. Multiple-group path analysis involves testing the path model with the parameters constrained (i.e., fixed to be the same) and unconstrained (i.e., allowed to be freely estimated) across the grouping variable (i.e., SOC). A Chi-square difference test was used to determine if there is a significant difference between the Chi-square estimates for the constrained and unconstrained models. If the Chi-square difference test indicated a statistically significant then the conclusion is that the strength of the path is statistically different for White students and SOC. All Chi-square difference tests were performed in Excel.

Results

Preliminary Analyses

Means, standard deviations, and bivariate correlations of the study variables are presented in Table 1 for White students and SOC. All five steps of the bystander intervention model were positively and significantly correlated with the other steps of the model for both White students and SOC. For SOC, empathy was positively and significantly correlated to Interpret, Accept, Know, and Act, but the correlation for Notice was not significant. Self-Efficacy was positively and significantly correlated with each step. Moral disengagement was negatively and significantly correlated with Notice, Interpret, Accept, and Act. For White students, empathy and self-efficacy were positively and significantly correlated with all five steps, and moral disengagement was negatively and significantly correlated with all five steps.

Table 1 Means, standard deviations, and bivariate correlations of main study variables for SOC and White students
Full size table

We also conducted a one-way ANOVA to compare the means of the two student groups (i.e., White and SOC) on moral engagement, empathy, and student engagement, which revealed a nonsignificant effect of students’ race on these variables.

Main Analysis

To answer the main research question, a multi-group path analysis was conducted. Sex was included as a covariate in the model since women of color may experience gendered microaggressions, which are unique to them for holding two marginalized and intersecting identities: woman and person of color (Lewis et al., 2017). However, since sex was not a main focus of the project, the paths are not interpreted. See Figs. 1 and 2 for a diagram and Table 2 for standardized and unstandardized path coefficients, standard errors, and p values for all paths. Overall, the results suggest a slightly different pattern of associations for White students and SOC across the three exogenous variables and each step of the bystander intervention model, but, in general, race was not significantly related to the variables.

Fig. 1
figure 1

Statistically significant paths for SOC participants. Note: Gray paths are not statistically significant. Refer to Table 2 for path coefficients

Full size image
Fig. 2
figure 2

Statistically significant paths for White participants. Note: Gray paths are not statistically significant. Refer to Table 2 for path coefficients

Full size image
Table 2 Unstandardized and standardized path coefficients, standard errors, p values, and χ2 difference analyses
Full size table

Moral Disengagement

Moral disengagement was significantly and negatively related to each step of the model for White students, but for SOC, it was only significantly associated with the Act. For White students, the association between moral disengagement and each step was negative and statistically significant (Notice, \(\beta\) = −0.28, p < 0.001; Interpret, \(\beta\) = −0.22, p = 0.007; Accept, \(\beta\) = −0.30, p < 0.001; Know, \(\beta\) = −0.28, p < 0.001; Act, \(\beta\) = −0.24, p = 0.003). For SOC, however, the association between moral disengagement and each step of the model was not statistically significant for any step (Notice, \(\beta\) = −0.02, p = 0.877; Interpret, \(\beta\) = −0.13, p = 0.37; Accept, \(\beta\) = −0.017, p = 0.896; Know, \(\beta\) = 0.11, p = 0.25; Act, \(\beta\) = 0.11, p = 0.098). The Chi-square difference test indicated that there was a significant Know step when comparing White students and SOC, with the path being stronger for White students.

Empathy

For White students, empathy was significantly and positively associated with Interpret (\(\beta\) = 0.15, p = 0.044) and Act (\(\beta\) = 0.19, p < 0.001), but there was not a statistically significant association with Notice (\(\beta\) = −0.01, p = 0.837), Accept (\(\beta\) = 0.12, p < 0.053), or Know (\(\beta\) = 0.01, p = 0.868). For SOC, the association between empathy in each step of the model was not statistically significant (Notice, \(\beta\) = −0.05, p = 0.525; Interpret, \(\beta\) = 0.06, p = 0.38; Accept, \(\beta\) = 0.14, p = 0.061; Know, \(\beta\) = −0.07, p = 0.383), except the path between empathy and Act was positive and statistically significant, \(\beta\) = 0.16, p = 0.011). Though there was a different pattern in which variables were significantly related to the steps of the model, the Chi-square difference test indicated that there was not a statistically significant difference in the strengths of the paths when comparing White students and SOC. Hence, race was not a significant moderator.

Self-Efficacy

Self-efficacy was related to each step of the model for both White students and SOC. For White students, the association between self-efficacy and each step of the model was positive and statistically significant (Notice, \(\beta\) = 0.24, p < 0.001; Interpret, \(\beta\) = 0.39, p < 0.001; Accept, \(\beta\) = 0.18, p = 0.012; Know, \(\beta\) = 0.31, p < 0.001; Act, \(\beta\) =0.43, p < 0.001). The same pattern emerged for SOC (Notice, \(\beta\) = 0.43, p < 0.001; Interpret, \(\beta\) = 0.49, p < 0.001; Accept, \(\beta\) = 0.41, p < 0.001; Know, \(\beta\) = 0.69, p < 0.001; Act, \(\beta\) = 0.63, p < 0.001).Similarly, the Chi-square difference test indicated that there was not a statistically significant difference in the strengths of the paths when comparing White students and SOV, except there was a significant difference for the path with Know, with the path being stronger for SOC.

Discussion

In this study, we used microintervention strategies (Sue et al., 2019) and the bystander intervention model (Latané & Darley, 1970) to conceptualize bystander intervention in racial microaggressions in a sample of diverse college students (White students and SOC). Informed by the bullying literature (Allison & Bussey, 2016), we also selected and examined potential social cognitive predictors (i.e., moral disengagement, empathy, and self-efficacy) of each of the five steps of the bystander intervention model (i.e., Notice, Interpret, Accept, Know, and Act) to understand how these predictors may be associated with engagement in bystander intervention. This study fills a much-needed gap in the literature on intervening in racial microaggressions by not only utilizing a model that explains bystander intervention in racial microaggressions and examining these differences by race, but also elucidating social cognitive predictors of bystander intervention. These findings can inform the development of interventions aimed at reducing of racial microaggressions on college campuses and has implications for fostering allyship, civil courage, and anti-racist behaviors in bystanders.

Moral Disengagement

We hypothesized that moral disengagement would be negatively associated with each step of the bystander interventions model for White students and SOC. This hypothesis was fully supported for White students, but not for SOC, despite mean score differences on this measure being nonsignificant (M = 9.79 versus M = 9.92). More specifically, moral disengagement was significantly and negatively related to each step of the model for White students. For SOC participants, it was not significantly associated with any step of the model. Moreover, our results also showed that the Know path was statistically stronger for White students than for SOC, so race moderated this path only. Moral disengagement is the process by which White students may defend immoral actions, by rationalizing that their moral standards do not apply in specific situations (Bandura, 1990). Bandura (1990) identified several mechanisms for how these behaviors are excused, such as cognitive restructuring or downplaying their responsibility or effect. White students may engage in the displacement of responsibility where they do not feel it is their role to intervene when they witness a racial microaggression. They may also minimize the effects of their behaviors by understating the effect of racial microaggression for targets. Thus, White bystanders who can morally disengage may excuse their inaction by using these mechanisms to reason that their intervention is not needed. Hence, creating interventions targeted at raising awareness about what racial microaggressions are and the negative effects (Nadal et al., 2014) they can have on SOC seems critical to increasing moral engagement. Such interventions may humanize SOC and build empathy in White students.

For SOC, once both empathy and self-efficacy were included in the model, moral disengagement was not significant—the variance was explained by empathy and self-efficacy. Moral disengagement may be less relevant in bystander intervention for SOC given the salience of an SOC’s racial identity and the inability of SOC to separate race from their lived experience. The nonsignificant relationship between moral disengagement and the steps of the bystander intervention model may also suggest additional important mediators at play such as social norms around intervening or not (e.g., concerns around intervening making a situation worse) and other aspects of the cost-reward model (Dovidio et al, 2016).

Empathy

We hypothesized that empathy would be positively associated with each step of the bystander intervention model. This hypothesis was partially supported. For SOC, increases in empathy was positively linked to choosing to intervene (i.e., Act). SOC may be more inclined to feel an emotional connection to the target of a racial microaggressions because of their personal experience of being a target (Sue et al., 2007, 2019). Because of the emotional response they have witnessing a racial microaggression, they may “jump in” or act before progressing through each of the steps described by Latané and Darley (1970). For White students, empathy was significantly and positively associated with Interpret and Act suggesting that higher levels of empathy was associating with interpreting an action as a racial microaggressions and choosing to act. Perhaps the preceding steps for White students are linked more closely to empathy because there may be cognitive dissonance over whether to Act and intervene as a bystander. As noted by William et al. (2023), intervening as a White person involves some risk of avoidable consequences from their same race peers.

Self-Efficacy to Intervene

We hypothesized that self-efficacy to intervene would be positively associated with each step of the bystander intervention model. Our hypothesis was fully supported. For both SOC and White students increases in self-efficacy or confidence in how to intervene was significantly positively associated with Notice, Interpret, Accept, Know, and Act. It makes sense that the more comfortable or confident one feels about intervening, the more likely they will be to intervene. This finding is supported in the bullying literature (Fredrick et al., 2020; Ma et al., 2019; Menolascino & Jenkins, 2018), and with this study has been expanded to bystander intervention in racial microaggressions. Of all the social predictors examined, self-efficacy was the most predictive of engaging in the five steps of bystander intervention in racial microaggressions for both White students and SOC. These results may allude to students wanting to intervene but choosing not to because they fear intervening in ways that put themselves at risk and fail to improve the situation for the target. These findings highlight a potential point of intervention to develop bystander intervention in racial microaggressions could be to increase the bystander’s self-efficacy to intervene, which can be increased by education for bystanders about racial microaggressions, how to intervene effectively, and practice (Sue et al., 2019). It is important to note that self-efficacy to intervene may be increased in different ways, depending on the race of the bystander and the perpetrator, and that these racial differences in actors will need to be considered in bystander intervention training opportunities. Finally, given the lived experiences of racial microaggressions for SOC, it is not surprising that the Know path was stronger for SOC than White students; SOC are personally aware of racial microaggressions and of their negative impacts. Hence, they may be able to take on the perspective of the target and have preferences around how they would want to a bystander to intervene if the roles were reversed.

Strengths, Limitations, and Opportunities for Future Research

The current study has numerous strengths as the first to empirically apply the bystander intervention model to racial microaggressions. First, we used Sue and colleagues’ (2019) microintervention strategies to adapt an existing bullying intervention measure to create the Bystander Intervention Measure for Racial Microaggressions. Second, we used the bystander intervention model (Latané & Darley, 1970) to conceptualize bystander intervention in racial microaggressions. Third, we used social cognitive theory (Bandura, 1986) to inform the selection of social cognitive predictors that may be associated with each step of the bystander intervention model among White students and SOC.

The study, however, is not without limitations. First, although our overall sample size was a strength, ideally, we would not want to group all SOC in to one group but rather to explore the diversity among this population by focusing more closely on the nuanced experiences of their specific identities. For example, Black college students may have different experiences of racial microaggressions than Asian college students so we should examine these groups differently. Future researchers should attempt to increase SOC sample sizes (for example, by oversampling SOC) to facilitate this nuanced analysis and potential replication and extension of findings. Second, this study is one of the first investigations applying the five steps to bystander intervention in racial microaggressions, so additional work is needed to ascertain the range of contextual and trait characteristics that influence intervention decisions. Future work should include qualitative work and other types of quantitative work (e.g., vignettes) to gather additional information about what influences whether and how individuals will intervene. Third, the sequential nature of the model should also be accounted for in future investigations. In this study we isolated each step of the model and examined predictors of each step in isolation. However, we know that intervention may not happen in stages and be more nuanced and dynamic. Fourth, we examined sex as a covariate, but future investigations could include other variables that may related to the constructs in the study, such as recruitment type, socioeconomic status, political affiliations, etc. Not only would this information help provide a richer description of the sample, but also these characteristics may be predictive of the relationships among these variables. Fifth, for this study, we recruited our sample from two different sources (online sampling via Mturk) and college students and did not look at differences by recruitment type. Future research should examine differences between these two sampling methods. Finally, we only examined racial microaggressions in this study when many types of microaggressions may be experiences by the same person. Future research should include consideration for intersectionality (e.g., gendered racial microaggressions) and how when a target has multiple marginalized identities may impact bystander intervention. Furthermore, questions targeting the ways in which self-efficacy to intervene can be increased based on racial backgrounds should be explored. Since differing risks exist based on racial backgrounds of the perpetrator, target, and bystander, the roles these play and how to navigate these complex racial situations effectively needs to be examined.

Implications

Prevention efforts to address racial microaggressions on college campuses should capitalize on bystander intervention’s role in reducing these acts. Trainings should be developed around reducing moral disengagement, as well as increasing empathy and self-efficacy to intervene. Self-efficacy to intervene appears to be the most important across racial groups and should be tailored using different racial scenarios. Bystander intervention in racial microaggressions is an example of civil courage (i.e., courageous behavior that is accompanied by indignation about injustice that is intended to exemplify or transform societal and ethical norms without singularly focusing on the social cost to oneself (Greitemeyer et al., 2007; Williams et al., 2023). Civil courage is a prerequisite for social justice orientation, which embraces the fair treatment of SOC and precipitates fair opportunities and outcomes for all college students (Hochman & Suyemoto, 2020). It is important to note that social justice orientation goes above and beyond the absence of discrimination but also the development of intentional systems and supports to develop and maintain equity (Hochman & Suyemoto, 2020). Although an important act of individual intervention, bystander intervention should not be substituted for making systemic changes that reduce racism and its effects for SOC on college campuses.

Civil courage is a prerequisite to allyship (Williams et al., 2023). Preparing antiracists means educating all students on the injustice inherent in societal rules and promoting a willingness to break them (Williams et al., 2023). Indeed, Williams and colleagues (2023) include bystander intervention (i.e., defending convictions in a hostile forum) in their list of ten recommendations for fostering civil courage. However, the context is an important consideration when engaging in bystander intervention. As acknowledged by Sue and colleagues (2019), both White students and SOC need to: (1) pick their battles, (2) consider where and when they choose to address the offender, (3) adjust the response as the situation warrants, (4) be aware of relationship factors and dynamics with perpetrators, and (5) consider the consequences the mincrointervention (i.e., intervening) especially when power differentials exist between the target and the perpetrator. Such considerations should be discussed in designing interventions in practice settings and in training opportunities in programs. However, sometimes risk and causing “good trouble” is needed to facilitate change so lessons around navigating different contexts within the contexts of personal risk should also be discussed. Bystanders should have a system of supports to engage in bystander intervention, especially for SOC. Allyship takes work.

Conclusion

The five-step bystander intervention model (i.e., Notice, Interpret, Accept, Know, and Act) is a useful way to incorporate Sue et al.’s (2019) microintervention strategies to address racial microaggressions on college campuses. There are differences among SOC and White students in the utility of moral disengagement, empathy, and self-efficacy to predict bystander intervention steps highlighting the need for culturally tailored bystander intervention training to facilitate engagement in bystander intervention to address racial microaggressions. Of the three, self-efficacy seems to be the most predictive of engagement in the five steps. Practitioners and researchers should also be aware of bystander intervention in racial microaggressions and the ways in which they should advocate for SOC to address racial microaggressions.