Modifiers of Neighbors' Bystander Intervention in Intimate Partner Violence: A Concept Mapping Study

Abstract

Encouraging bystander intervention in intimate partner violence (IPV) against women is potentially an important method of reducing the prevalence of such violence in urban communities. Most existing research has been conducted on campuses and in relation to sexual violence among teens or young adults. Our understanding of which bystander behaviors are feasible is nascent, and our knowledge of which situational factors influence neighbors' self-reported willingness to intervene is underdeveloped. We conducted a concept mapping study to identify potential bystander intervention behaviors in IPV among neighbors in urban settings; we also assessed whether perceived feasibility and effectiveness of those behaviors varied by situational characteristics. Using data collected from 41 residents of a low-income New York City neighborhood in late 2011, concept mapping was used to create a conceptual map of the 74 behaviors identified by participants. We examined participant differences in mean feasibility (i.e., that the participants “could” or “would” enact a behavior), feasibility given two situational characteristics (if the couple was perceived to have a history of IPV, and if children were believed to be involved or present), and perceived effectiveness of bystander behaviors. Differences across select sociodemographic factors of participants were also analyzed. A 13-cluster solution emerged, with clusters of bystander behaviors grouped into four larger cluster areas: victim focused, parenting/education focused, perpetrator focused, and community involvement focused. Bivariate analyses revealed that participants rated the four cluster areas as more feasible when a child was believed to be involved. Male participants rated intervention as less feasible when the couple was believed to have a history of IPV. Participants who reported a history of IPV victimization rated all four cluster areas as less effective on average, as compared with participants without a history of IPV. This study explored bystander intervention into IPV outside of a college context and among urban adults living in high-poverty areas. Results suggest that the presence of children and perceived history of IPV may affect bystander intervention. Specific recommendations to build the research base on bystander intervention in adult IPV as well as what situational, sociodemographic, and other factors mitigate against intervention among potential responders are offered.

Introduction

Despite decades of grass-roots action and legal and policy responses, intimate partner violence (IPV) against women continues to be a pervasive problem both globally and in the United States (Campbell 2002; Stockl et al. 2013). The health effects of IPV are well-documented, with lasting consequences to physical, mental, sexual, and social health among victims, including posttraumatic stress, gynecological complications, chronic pain, and depression (Campbell 2002; Ellsberg et al. 2008). A public health approach to the issue extends our understanding of IPV beyond the individual level to the situational, community, and socioenvironmental factors involved.

The body of literature examining the situations and processes by which IPV victims cope with and attain safety is growing, as is our understanding of how communities can prevent IPV. Studies on victim help-seeking have consistently revealed that IPV victims access formal actors (e.g., police, social or mental health services) less often than informal actors (e.g., family, friends, and neighbors) (Liang et al. 2005; Simmons et al. 2011; McCart et al. 2013). Informal actors are typically easily accessed, are proximal, and may be able to intervene before, during, and after IPV events. In addition, informal actors are often socially interactive co-creators of the local social norms that may influence IPV and bystander behavior, thus the current focus on bystanders in IPV and sexual violence prevention research and practice (Banyard 2008; Beeble et al. 2008; Casey and Ohler 2012; McMahon and Banyard 2012). However, the majority of this research has focused on young adults living on campus in college settings rather than on adults living in community or neighborhood settings. And although the literature is growing on how community characteristics may contribute to the prevalence of IPV and factors that influence survivor help-seeking, little has examined how community members intervene (Beeble et al. 2008) or what contextual factors may modify such behavior (Frye et al. 2012).

To address this gap in the literature we conducted a concept mapping study of potential IPV bystander interventions among adults (neighbors) living in a low-income, urban neighborhood. We used concept mapping to address two research aims: first, to describe bystander strategies generated by members of a New York City (NYC) neighborhood, and, second, to examine the effect of specific situation-level characteristics hypothesized to mitigate the self-reported feasibility and perceived effectiveness of the identified intervention behaviors. Concept mapping is a mixed-method approach designed to provide a visual representation of how participants conceptualize a specific phenomenon; understanding how lay people conceptualize bystander intervention is an effective formative research method for what is a complex social and health problem. Previous studies have utilized concept mapping to examine neighborhood influences on the prevalence and severity of IPV (O'Campo et al. 2005), and to generate community-identified bystander strategies (Frye et al. 2012). The current study builds upon this work with an examination of mitigating factors to bystander intervention.

Theoretical framework

This study conceptualizes IPV through both feminist and socioecological theoretical lenses. Feminist theories posit that IPV represents an extreme manifestation of a system of oppression that produces and maintains gender-based inequality and stratification (Stark 2007). A socioecological perspective frames IPV as behavior that is influenced by factors at various levels of the social ecology, including the individual (e.g., perpetrator factors), interpersonal (e.g., relationship dynamics), institutional (e.g., criminal legal policies), and broader structural and social factors (e.g., political economic systems). The theory of gender and power (Connell 1987; Wingood et al. 2000) posits that gender inequality is produced, and violence against women justified, through the sexual division of labor and power, as well as social norms that justify and reinforce these structures. Combined, these theoretical frameworks allow us to consider various social actors (IPV victims and neighbors, in this study) and interactions across a range of ecological levels. Empirical evidence supports the notion that institutions and norms influence IPV prevalence. For example, research continues to demonstrate that public perceptions of survivors remain low (Finkelhor et al. 2001; Waltermauer 2012) and that police continue to perceive reports of violence among couples as less serious than other forms of crime (Weller et al. 2013).

Bystander intervention behavior

Bystander intervention behavior is conceptualized as a cognitive and affective decision-making process influenced by individual, situational, and contextual factors (Banyard et al. 2004). More specifically, bystander interventions range from intervening in discrete acts of violence, speaking out or making public commitments toward social change on a social or public health issue, and building skills for a range of prosocial behaviors (Banyard et al. 2004, 2009). Individual-level factors that influence intervention in generic violence situations have been well-studied (Fischer et al. 2011; McMahon and Banyard 2012). Germinal research by Shotland and Straw (1976) showed that men are more likely to intervene in physical assaults when both actors are male and when it is a high-danger situation, as compared with situations that are perceived to be IPV (Shotland and Straw 1976). Specific to IPV, Frye (2007) found that younger age, stronger feelings of self-efficacy to intervene, and less tolerant attitudes toward IPV were associated with self-reported likelihood of intervention among adults (Frye 2007). Beeble and colleagues (2008) found that gender (women) and younger age were associated with intervention and that participants who reported previous IPV victimization were more likely to offer tangible support (e.g., giving a ride, lending money), but not emotional support or help to enact formal systems, to IPV victims (Beeble et al. 2008). Banyard's collective research identified the following correlates of bystander intervention in sexual violence situations: sex/gender, knowledge of the issue, bystander attitudes (Banyard 2008), age, feelings of responsibility, decisional balance (pros outweighed cons), and self-efficacy (Banyard and Moynihan 2011). Banyard and colleagues (2009) underscore that perceived effectiveness and self-efficacy to perform a behavior are key to engaging community members in bystander interventions. These individual-level factors are wrapped up in the broader social and environmental context (e.g., availability of community resources, experiences with formal responders to violence such as police) as well as the situational context (e.g., where and how an act of violence occurs or awareness of IPV arises, who else is involved in or witness to the event) (Banyard et al. 2009).

Contextual factors and bystander behavior

There is little research on how bystander behavior occurs among adults living in urban neighborhoods, with most studies conducted in college settings and analyzing young adult bystander attitudes, intentions, and behaviors (Beeble et al. 2008; McDonnell et al. 2011). In order to promote community-level IPV prevention interventions among urban adults, we must consider how contextual or situational factors impact people's perceptions of IPV and mitigate against their likelihood of intervention. Prior victimization within the couple and the presence of children are two factors that have been found to affect formal interventions (e.g., police response) and may also influence the responses of informal actors (i.e., friends, family, neighbors, or other bystanders) (DeJong et al. 2008; Davies and Roger 2009; Weller et al. 2013). Reluctance to intervene if the couple has a history of IPV may reflect a sense of futility of trying to help women who appear to “choose” to stay with or return to abusive men. Evidence suggests that negative attitudes toward victims are stronger when a victim returns to the abuser (Yamawaki et al. 2012). And while it is well documented that the presence of children in the home both inhibits and motivates victim help-seeking (Finkelhor et al. 2001; Kelly 2009; Swanston et al. 2013), less work has examined how it influences bystander behavior (Fledderjohann and Johnson 2012). Waltermaurer's review found consistent and strong justification for and tolerance of IPV when actors suspect neglect or poor mothering of a child by the victim (Waltermauer 2012). Other qualitative work reveals that people view women who experience IPV less as “victims” and more as “bad mothers” when there are children involved (Kelly 2009). It is reasonable to posit that the presence of children in an IPV situation may affect bystander behavior, but the direction of such behavior remains unclear. Given this evidence, “history of IPV” and “presence of a child” were selected as key mitigating factors in bystanders' perceptions and decisions to act in IPV.

In order to design programs that support informal actors in safe and effective bystander intervention behavior, we first need to develop our knowledge base about the victim, as well as the situational and contextual factors that influence self-reported bystander intervention (or lack thereof) in IPV situations. In our prior research using concept mapping, we assessed bystander behaviors that individuals might enact to prevent IPV among an adult sample in a different NYC borough and neighborhoods, and the role of informal social control/social cohesion (Frye et al. 2012). In this work, we described a much wider range of behaviors than had previously been described and measured; however, we were not able to examine whether select factors encouraged or discouraged the bystander behavior described by that sample. There remains little research to date that improves our understanding of what factors mitigate against enacting these bystander strategies among adult urban residents. This study fills this gap by using concept mapping, a mixed-method participatory approach to examine whether mitigating factors (e.g., the couple has a “history of IPV” or “a child is present” in the home) influence self-reported feasibility and effectiveness of engaging in bystander intervention behavior among urban adults.

Study Methods

Concept mapping overview

Concept mapping combines qualitative focus groups with statistical modeling to produce visual “maps” of a research topic. It is a multistep method that results in an understanding of how individuals conceptualize a particular topic or the focal phenomenon (Trochim and Kane 2005; Kane and Trochim 2007). First, a brainstorming group (or series of groups) is held with community members to generate ideas or strategies addressing a research question. Strategies generated by participants are then reviewed by the research team for clarification, and presented back to another focus group (or series of groups) for sorting and rating. Sorting the statements involves every study participant organizing the statements into piles they determine to be conceptually similar and labeling each pile. In a separate exercise during the same session, study participants rate each statement according to a predetermined quality or characteristic (in this case, feasibility and effectiveness) on a 5-point Likert scale in relation to the other statements. Finally, the research team uses the Concept Systems software for data entry, analysis, and modeling to produce visual maps that demonstrate the relationship between sorts and ratings across participants.

Sampling methods and study procedures

In this study, we recruited 41 residents of a low-income NYC neighborhood using street-intercept methods to participate in a concept mapping study on “domestic violence against women living in the neighborhood.” For both brainstorming and sort-rate groups, eligible participants had to be 18 or older, able to read and write in English, and residents of the focal neighborhood for at least 1 year (determined by self-report and identifying their approximate residence on a map). Those screened eligible were given a card with information about the upcoming session, study contact number, time and location of the session, and exceptions to confidentiality, including disclosure of intent to harm themselves or others and child maltreatment or abuse. Before the groups, all participants engaged in the informed consent process. Participants of both the brainstorming and sort and rate groups received $40 and a round-trip subway/bus fare for their time and travel. The study was reviewed and approved by the Institutional Review Board (IRB) of the New York Blood Center.

Three brainstorming groups (N = 28) were convened in late 2011; participants first completed a brief demographic survey, which, in addition to age, race/ethnicity, education, U.S.-born status, etc., also assessed whether the participant had ever been a victim of or perpetrated IPV. Next, domestic violence was defined^* and participants were asked to generate bystander intervention strategies by responding to the following question: “One specific action a neighbor or group of neighbors could do to prevent domestic violence is ________?” Afterward, the research team reduced the original list of 92 statements (intervention strategies) to 74, reviewing for clarity and removing redundancies. Participants of the brainstorming groups were invited to return for the sort and rate process; additional participants were recruited in the same manner described above, resulting in 41 participants participating in a series of sort and rate groups. Participants were asked to sort the 74 statements into piles they felt had a similar theme, and label each pile (these sorted groups and labels are used later to form the concept map clusters). They then rated the statement (intervention strategy) in terms of general feasibility and perceived effectiveness, and again for the statement's feasibility given two mitigating factors. General feasibility was defined as: “How feasible (you could and would do this action) would it be for you?” Effectiveness was defined as: “How effective (or useful) do you see this action being in the prevention of domestic violence?” Feasibility to prevent IPV by enacting the behavior(s) generated was also assessed given two situational characteristics. The first was feasibility given the perception of a “history of IPV,” defined for participants with examples such as “the couple breaks up and gets back together often” or “the couple has been together for a very long time” or “the woman stays with the man, or defends him, even though he is abusive to her.” The second was feasibility when “a child is present,” which was defined as a situation where there are “kids living within the apartment” or “kids are at least sometimes involved with, or witness, the abuse.” Participants scored each statement relative to the other statements using a 5-point Likert scale, with values ranging from 1, “not at all feasible/effective,” to 5, “extremely feasible/effective.”

Concept mapping analysis

There are three data sources in concept mapping: the statements generated by participants; the piles into which participants sorted the statements; and the ratings participants attach to statements for feasibility and perceived effectiveness. We used Concept Mapping statistical software (Concept Systems Inc., 2003) to create the cluster maps. PASW/SPSS 18.0 (Chicago, IL) was used to conduct independent and paired t-tests to examine differences in the two mitigating factors and select sociodemographic factors, including gender, education, and lifetime victimization. Concept mapping uses multidimensional scaling (MDS) to create a visual map of how the statements relate to each other using the pile sorts. In MDS, each statement becomes a point on a map and a summed square similarity matrix table is represented as distances in Euclidian space (Kruskal and Wish, 1978). The closer the statements are to each other, the more likely they were to be sorted in the same piles by participants. Next, agglomerative cluster analysis, using Ward's algorithm, partitions the MDS map hierarchically into nonoverlapping clusters that represent sorts or ideas. The “final solution” is determined by assessing the bridging values (range 0–1) between statements as the number of clusters increases. Bridging values indicate whether a statement was sorted with other statements that are close to it or farther away on the map (Kane and Trochim 2007); lower bridging values suggest a closer relationship with other statements in the same cluster. For example, as the cluster solution increases from 8 to 9 distinct clusters, a bridging value of 0.80 may split into two clusters with lower values, indicating that some statements in the original cluster were more likely part of a separate, distinct cluster. At times, the bridging value for only one of the new clusters will decrease and the analyst assesses the utility of the split, based on the new and old bridging values as well as looking subjectively at the contents of the two new clusters. Kruskal's stress statistic, a sum-of-squared differences estimator, measures the accuracy of this relationship; values lower than 0.30 are considered to be adequate in concept mapping (Kane and Trochim 2007). The final cluster arrangement is superimposed onto the point map and each cluster is named or described.

Results

Participant characteristics

The sample was 54% male and participants' age ranged from 18–53, with a mean age of 33 (SD = 8.6). The majority (58%) identified as black/African American or Hispanic/Latino (44%); 32% identified as mixed race/other. Most (88%) were born in the United States. Twenty-seven percent of the participants had less than a high school diploma and the majority were unemployed (76%). Twenty-two percent reported ever perpetrating IPV and 41% reported being victimized (Table 1).

Table 1.

Selected Sociodemographics (N = 41)

Sociodemographic	N	%a
Male	22	54
Mean age (SD)	33 (8.6)	—
Race
Black/African American	24	58
Mixed/other	13	32
White	1	2
Hispanic/Latino	18	44
U.S. born	36	88
Less than high school diploma/GED	11	27
Unemployed	31	76
IPV perpetration, lifetime	9	22
IPV victimization, lifetime	17	41

^aPercentages may not total to 100% because of missing data.

IPV, intimate partner violence; SD, standard deviation.

Cluster solution

As described above, the cluster solution used in analysis was determined by assessing the bridging values between statements as the number of clusters increased. For this study, 74 statements were sorted by 41 study participants. Based on assessing bridging values as the cluster solution increased, the research team determined a final 13-cluster solution because bridging values did not improve with 12 or 14 compared to 13 clusters. This means that 13 clusters best represent the “sorts” of each participant. The stress statistic was 0.23, representing stability, and was similar to our findings in the previous study. When generating the cluster solutions, the software system uses the sort labels from participants to name each cluster. The research team renamed select clusters for clarification and based on knowledge of the pile content (Fig. 1).

FIG. 1.

Thirteen-cluster concept map (N = 41).

Just as data points on the map (statements) that are closer together are more likely to be sorted in the same cluster, the distance between clusters in the final solution also represents how closely they are related. For example, the “Getting Physically Involved” cluster is much closer to the “Involving 911” cluster than it is to “Parenting (mother) Strategies,” meaning the contents of the first two “sorts” are more alike. This makes sense given that physical intervention and calling 911 would likely occur during a physical assault or moment of crisis, whereas providing parenting support or advice reflect primary prevention strategies; each involve distinct decision-making processes and tactics.

The research team then used the cluster replay analysis to identify larger, overarching themes of the 13-cluster solution. Cluster replay analysis records the changes in cluster solution from the point where each statement represents its own cluster (74 in this case) to one single cluster. Researchers and participants can see which clusters merge to create larger clusters, and can therefore blend qualitative interpretation of the cluster solution with the statistical representation of the map. Using this process, we identified that the 13 clusters fit into 4 larger “cluster areas” with key organizing themes: 1) victim focused; 2) parenting/education focused; 3) perpetrator/formal systems focused, and 4) community involvement focused (Fig. 2).

FIG. 2.

Four-area concept map (N = 41).

Responses for the victim-focused cluster included “refer the woman the hospital/services/counseling,” “have a girl's day with friends or neighborhood women,” and “serve as a ‘healthy couple’ example in your own relationship.” The parenting/education area included actions such as “have workshops or meetings at public schools about domestic violence,” “moms and kids talk about roles of boys and men in the home and with partners,” and “parents to be more strict with children.” The perpetrator/formal systems cluster included “confront the man before/after the incident,” “bang on ceiling or pipes,” and two items related to enlisting formal systems, “call the police” and “anonymously call 911”; the community focus cluster included actions like “leave a note at the door,” “shame silent bystanders,” or “promote a positive atmosphere in your building.” Similar, but not identical, areas were identified in our previous study (systems focused, perpetrator focused, victim focused, and community focused) and novel responses were identified in this neighborhood (Frye et al. 2012). For example, in the current study, participant-generated strategies for victim and parenting (with a focus on mothers) responses were grouped into separate cluster areas, whereas these strategies were grouped into the same cluster area in the 2012 study. Similarly, the 2012 study revealed a greater distinction between perpetrator- and systems-focused responses, evidenced by the emergence of unique cluster areas. As discussed earlier, bystander behavior is affected in part by the neighborhood social environment. People in different neighborhoods have different experiences and perspectives of their neighbors, community problems, police and other formal systems, as well as different access to formal and informal mechanisms for help. This may help explain why broad approaches to bystander intervention were similar across studies and samples, yet unique specific strategies emerged from each. Implications of this, as they relate to this study's findings more broadly, are discussed below.

Feasibility and effectiveness by cluster area

Looking across the four areas, study participants rated parent/education-focused strategies as the most feasible (3.53 [SD = 1.22], and perpetrator/systems as the least feasible (2.88 [SD = 1.08]). When rating each area on its potential effectiveness, participants rated parent-, community-, and victim-focused strategies as mid to highly “moderately effective” (3.87 [SD = 0.76]), and perpetrator-focused strategies lower, but still “moderately effective” (3.13 [SD = 0.92]). Average feasibility and effectiveness ratings never neared “extremely” feasible or effective.

Differences by respondent characteristics

On average, across all strategies listed, women rated strategies to intervene in IPV as more feasible than men (x = 3.57 vs. 2.89; t-statistic = 2.19, p < .05). Participants not born in the United States rated general feasibility as significantly higher than U.S.-born participants (x = 3.86 vs. 3.11; t-statistic = 3.39, p < .01). Participants who reported ever being a victim of IPV rated general feasibility lower than those reporting never being victimized, but these were not statistically significant. However, they did rate the general effectiveness of all strategies significantly lower than nonvictims (x = 3.18 vs. 3.75; t-statistic = 2.59, p < .05). This was also true for parent-focused cluster area (3.59 vs. 4.10; t = 2.19, p < .05), and community-focused (2.96 vs. 3.71; t = 2.94, p < .01) and perpetrator-focused (2.71 vs. 3.44; t = 2.63, p < .05) strategies as compared with nonvictim. Those with lower education (less than high school) also rated general effectiveness lower than those with a GED or higher education (3.08 vs. 3.65; t = −2.34, p < 0.05) (Table 2).

Table 2.

Feasibility and Effectiveness by Cluster (N = 41)

	Feasibility		Effectiveness
Cluster	Mean (SD)	Range	Mean (SD)	Range
Victim focused	3.33 (1.21)	1–4.96	3.59 (0.91)	1.25–5.00
Parenting/education focused	3.53 (1.22)	1–5.00	3.87 (0.76)	1.64–5.00
Perpetrator focused	2.88 (1.08)	1–4.76	3.13 (0.92)	1.04–5.00
Community involvement focused	3.03 (1.13)	1–4.83	3.38 (0.86)	1.58–5.00

Feasibility given mitigating factors

On average, participants rated feasibility to act when a child is present significantly higher than general feasibility (3.89 vs. 3.20; t = −4.29, p < .01). Mean scores by cluster area were also higher for feasibility when a child is present, as compared with general feasibility: victim- and parent-focused strategies were rated as “very feasible” (4.00 [SD = 0.72] and 4.14 [SD = 0.72], respectively), community-focused strategies neared a “very feasible” score (3.81 [SD = 0.78]), and perpetrator-focused strategies were rated as “moderately feasible” (3.57 [SD = 0.90]) (Table 3). Participants with lower education levels (less than high school) rated feasibility when there was a child present significantly lower than those with HS/GED or higher (3.45 vs. 4.05; t = −2.67, p < .05). There were no differences by sex in feasibility when there was a child present.

Table 3.

Feasibility by Cluster, Given Situational Characteristics

	History of IPV in neighbor's home		If a child lives in neighbor's home
Cluster	Mean (SD)	Range	Mean (SD)	Range
Victim focused	3.17 (1.14)	1–4.96	4.00 (0.81)	1.36–5.00
Parenting/education focused	3.58 (0.96)	1–5.00	4.14 (0.72)	2.64–5.00
Perpetrator focused	2.85 (1.09)	1–4.76	3.57 (0.90)	1.64–5.00
Community involvement focused	3.08 (0.96)	1–4.83	3.81 (0.78)	2.00–5.00

There was no significant difference between general feasibility ratings and feasibility when the couple was perceived to have a history of IPV. Across areas, feasibility ratings were similar if there was a history of IPV, with parent-focused strategies rated as “moderately feasible” (3.58 [SD = 0.96]) and perpetrator-focused strategies rated as “not very feasible” (2.85 [SD = 1.09]). Women rated strategies as more feasible when the couple was perceived to have a history of IPV than did men (3.47 vs. 2.91; t = 2.09, p < .05). There were no other differences for when the couple was perceived to have a history of IPV based on respondent characteristics (Table 3).

Discussion

This study sought to identify community-generated strategies for bystander intervention to prevent IPV among urban adults and to assess whether situational characteristics such as a perceived history of IPV or presence of a child modify self-reported feasibility and effectiveness. To our knowledge, this is the first study to explore bystander intervention outside of a campus sexual assault setting and in the context of IPV among urban, low-income adults. Community members identified 74 potential strategies that fell into 13 clusters, grouped within 4 overarching cluster areas. In our previous research, the 78 behaviors generated fell into 17 clusters, nested within 4 similar, but not identical, cluster areas (Frye et al. 2012). In this study we found that strategies focused on the perpetrator were rated the lowest in terms of both feasibility and effectiveness, consistent with our previous research. Perpetrator-focused tactics also remained lower in feasibility given mitigating factors than other approaches. In contrast, the parenting/education cluster was rated most feasible and effective in preventing future IPV, regardless of situational characteristics; numerous opportunities to intervene via this approach were identified. This suggests that personal safety and/or perceptions of formal systems may influence decisions of whether to intervene in a crisis or get physically involved. In contrast, primary prevention strategies were viewed as more feasible. The recent NO MORE study results confirm that fear of intervening is an important barrier to bystander intervention (GFK Public Affairs Corporate Communications Group 2013). Given that neighbors may interact with an IPV victim in various ways, and not just during an acute event, considering how neighbors can engage in primary prevention is critical. Designing programs that highlight opportunities to intervene outside of “crisis” situations, such as education-based or direct support strategies, may be the most acceptable, feasible, and potentially effective in preventing future IPV. Further research is needed to understand how frequently and in what contexts opportunities to intervene are presented, as well as what actual behaviors community members engage in and their outcomes.

Our study found no significant differences between general feasibility of intervention and feasibility when the couple has a “history of IPV,” suggesting that a perceived history of abuse may not reduce a neighbor's perceived feelings of responsibility to intervene. This finding is somewhat counterintuitive given that a history of abuse may produce fatigue or apathy regarding the effectiveness of intervention by neighbors and other informal helpers, resulting in lowered likelihood to act. Weller and colleagues (2013) also suggest that people may believe that abusive behavior is acceptable between couples but not strangers, influencing their perceived obligation to intervene. On the other hand, McMahon and Banyard (2012) suggest that more frequent visibility of IPV may make bystanders more likely to act in contrast to single acts of violence (e.g., cases of acute sexual assault). We found that women viewed strategies more feasible than men when there was a history of IPV (there were no differences between victims and nonvictims, however). This suggests key gender differences in perceptions of bystander intervention given a history of IPV. In addition, the parent-focused strategies tended to focus on primary prevention with the mother/female victim as opposed to the father/male perpetrator, suggesting that outside of crisis situations informal actors may be more comfortable focusing efforts on female victims as opposed to male perpetrators. More research is needed to identify strategies that are acceptable to bystanders when the history of IPV and associated attitudes about survivors are present, in addition to when a perpetrator is perceived to be dangerous.

Participants rated feasibility to intervene significantly higher when a child was involved. Consistent with Fledderjohann and Johnson's framework, community members may feel that their responsibility to act, or the gravity of the situation, is much greater when a child is involved, whereas privacy and nonintervention norms persist when it is “just a couple” (Fledderjohann and Johnson 2012). If neighbors express a higher feasibility to intervene when a child is present, and yet the effectiveness of certain interventions remains low particularly among those who have experienced IPV, it is important to understand whether their strategies will be supportive of the victim, child, and the mother–child relationship alike. The parent and education-focused strategies generated by community members in our study placed much of the responsibility on the women in the household and the community, suggesting that gender norms may also influence how people conceptualize responsibility for bystander intervention strategies. Given the similar gendered focus of interventions for when there is a history of IPV and as well when a child is present, further research is needed to explore survivor-centered responses. This work can then inform public education campaigns and other community interventions to shift norms about IPV, particularly related to perceptions of the survivor's responsibility to prevent, change, or end abuse.

This study has several limitations. It was a small, exploratory study conducted with a convenience sample of mostly racial/ethnic minority residents living in a low-income NYC neighborhood and therefore may not be generalizable. This study also assessed self-reported predictions about bystander behaviors, not actual behavior. In addition, all measures of lifetime victimization and lifetime perpetration were self-reported. Although a definition of domestic violence was offered for the focal question of study, it is unclear how each participant exactly defined and reported IPV in their own lives. Finally, we were not able to get participant feedback on the interpretation of the maps or findings, a key feature of concept mapping. While we explored participants' reasoning for their statement ratings on feasibility and effectiveness in conversations during the conduct of the study, systematic research needs to document the rationale behind self-reported willingness to intervene in IPV situations, and engage stakeholders throughout the research, program design, and implementation process. However, the findings of the study advance our knowledge in important ways despite its limitations. Despite the small sample, statistical significance was found in some important areas. Although many bystander studies focus on the college campus population (Fabiano et al. 2003; Banyard et al. 2004; Potter et al. 2009; Casey and Ohler 2012), before this study few have used urban community-based samples, which is needed to understand how various factors influence bystander intervention by adults and in settings that differ dramatically from campus ones.

Conclusions

This study used concept mapping to, first, determine various community-generated response techniques to violence and, second, to assess whether mitigating factors modify self-reported evaluations of the feasibility and effectiveness of bystander intervention behaviors to prevent IPV among low-income, urban-dwelling adults. The findings support and extend the current body of literature on the importance of understanding the situational and environmental context of IPV, including mitigating factors that reduce the self-reported likelihood of engaging in bystander intervention. Taken together with our previous results, our research offers important considerations for future study and work on bystander intervention in urban neighborhoods. Specifically, there is a need to study and consider the roles of bystander personal experiences and characteristics (i.e., prior victimization, gender, and other socioeconomic factors); bystander intervention strategies may vary between urban neighborhoods; direct intervention with the perpetrator and/or formal systems is generally perceived as less feasible and less effective than other options; mitigating factors such as history of IPV may not reduce intervention, but may need to be considered when promoting a fuller spectrum of intervention options; and mitigating factors such as presence of a child may increase feasibility to act, but may also encourage ineffective or even harmful responses. We must be cognizant of whether a potential interventions meet the needs of survivors and their children.

We recommend a number of research priorities to further develop this knowledge base. Higher perceived feasibility to intervene when a child is involved suggests that people felt greater prosocial norms of intervention when the consequences of IPV were perceived to affect a child. Bystander research and practice should explore further the potential mitigating impact of the presence of children on bystander behavior, and whether it is sensitive to survivor safety options and the mother–child relationship. Future research should explore this particularly among low-income adult, urban residents of high-IPV-prevalence areas and in conjunction with formal systems responses. And although we found concept mapping to be a useful approach to generating novel intervention strategies with community-based samples, more work is needed to determine which strategies are actually used and why. Research with adults living in urban areas and who have intervened is needed to describe the outcomes of the intervention and impact on subsequent IPV victimization and bystander behavior. Similarly, more research is needed on whether and to what extent a history of IPV impacts intentions to intervene. If a history of IPV, as suggested in our study and by others (McMahon and Banyard 2012), facilitates a fuller spectrum of bystander intervention strategies, more research is also needed to identify and evaluate the range of strategies. Research with community members should further explore reasons why formal responses such as calling 911 are viewed less favorably. To this end, sampling professionals from the formal sectors is needed to provide insights that may enhance responsiveness and to create dialog with community members to identify barriers and relevant strategies for IPV prevention and intervention.

Given the formative nature of the study, our findings call for more research in these areas in order to develop concrete programmatic and policy recommendations. We are thus reluctant to offer specific practice recommendations; however, the results do reflect the value in community participation to identify unique and diverse strategies and should be considered when planning or tailoring interventions. In particular, the similar cluster areas yet unique specific strategies generated by participants from our two studies suggest that norms (e.g., about IPV or women), experiences with and perceptions of available strategies (e.g., police vs. community education), and the actual resource structure of urban neighborhoods likely influence bystander behavior and ought to be assessed before implementing any one intervention. Further, based on identified strategies and discussion with participants during the study, urban adults understand the complexity of IPV and the need for multilevel and readily available strategies. Survivors should also be included in this work, to help match community-identified intervention strategies with survivors' self-defined needs. Based on our experience, concept mapping or other participatory methods should be employed in bystander intervention program design.

The need to identify bystander strategies that match survivors' needs and reflect the community context is clear. While community members generated the strategies for bystander intervention, ratings on both feasibility and effectiveness scales were not high. This was particularly true for respondents who reported IPV victimization. This suggests a few things: survivors are intimately familiar with the limitations of available interventions, strategies must be examined from a multilevel and multistrategy approach, and more efforts are needed to identify realistic, safe, and effective bystander intervention behaviors to prevent IPV. Given the prevailing rates of IPV in our society despite decades of effort, more research into community response to violence, beyond college settings, is needed to enact meaningful and community-based bystander intervention as well as inform policy for a greater impact in the movement to end violence against women.

Acknowledgments

This analysis was supported by a National Institute on Drug Abuse (NIDA) National Institutes of Health career development award to V.F. (DA-K01-020774). We thank the research participants for their effort in this study.

Author Disclosure Statement

The authors have no competing interests.