In 1995, Michael P. Johnson addressed the longstanding debate among scholars regarding gender a/symmetry in intimate partner violence (IPV), proposing that there are two distinct types of IPV represented in the literature that had been erroneously conflated under the umbrella term “domestic violence.” In particular, Johnson (1995) argued that situational couple violence is perpetrated at comparable rates by men and women and is most commonly captured in the work of family violence scholars, whereas intimate terrorism is perpetrated disproportionately by men against women in heterosexual relationships and is most commonly captured in the work of feminist scholars.Footnote 1 Relatedly, Johnson suggested that such differences corresponded to different methodological choices among feminist and family violence scholars regarding recruitment, measurement, and analysis. Johnson’s work has since evolved into an expanded typology of violence rooted in several key assumptions and with four distinct types of IPV: situational couple violence, intimate terrorism, violent resistance, and mutual violent control (Johnson, 2008, 2010, 2011, 2017). Given that IPV research has significant implications for policy and practice related to prevention and survivor safety and support, Johnson (2008) urged scholars “to demand that every piece of research on IPV be planned so that distinctions among types of violence are built into the research design” (p. 86). However, the literature on Johnson’s typology continues to be characterized by the scholarly divide it was intended to reconcile, with empirical support for Johnson’s typology (e.g., Conroy & Crowley, 2021; Haselschwerdt et al., 2019; Leone et al., 2010) existing alongside research that challenges Johnson’s typology, particularly regarding the role of gender in IPV (e.g., Dutton, 2006; Hines & Douglas, 2010; Straus, 2012).

In this paper, we systematically review the scholarly literature in order to assess the state of empirical research on Johnson’s typology of violence and to synthesize findings across research studies. In doing so, we aim to provide a clearer picture of the data that have supported or challenged the distinct types of violence in Johnson’s typology since its inception. Before reviewing and evaluating the literature, however, we offer a few clarifications about the scope of our paper and suppositions that guide our work. First, the first and second authors are feminist scholars of interpersonal power, control, and violence, and our scholarship includes but is not limited to testing Johnson’s typology. Conroy and Crowley also have professional experience as victim advocates and partnering with community organizations providing services to survivors. These experiences have motivated us to study the potential utility – and limits – of Johnson’s typology as we critically consume the extant literature testing his typology, design our own research studies, and engage with community partners. We also believe that empirical data collected using rigorous research methods are necessary for evaluating the utility and validity of Johnson’s typology in research and practice; we do not accept his typology at face value.

Second, although Johnson offers a typology of violence, his 2008 book offers a detailed explanation of the process of developing the typology, clarification of key concepts, identification and explanation of a number of core assumptions, methodological recommendations, and the potential for validating the typology and its assumptions – all of which Boss (2016) deems characteristic of “a starting point for dialogue about the process of theory development” (p. 280). Johnson (2017) and others have also suggested that the typology is a theory of violence (e.g., Capaldi & Kim, 2007; Meier, 2015), and more generally that a typology is not a mere classification system but a “unique form of theory” in and of itself (Niknazar & Bourgault, 2017, p. 192). While it is beyond the scope of this paper to debate whether the typology is a theory, we argue that testing a typology is necessary for determining its theoretical and practical utility and validity.

Johnson’s Typology of Violence

Central to Johnson’s typology is the role of gender in IPV,Footnote 2 an issue that has long been debated in the IPV literature (Johnson, 2006, 2008). According to Johnson and other feminist scholars, men disproportionately perpetrate violence against women, and equal rates of perpetration does not mean that men and women experience IPV similarly (Anderson, 2005; Johnson, 2008). On the other hand, family violence scholars generally suggest that IPV is not about gender; violence is a relational problem perpetrated at equal rates between men and women (Hamel, 2009; Straus, 2012). Johnson (1995) first attempted to settle this debate by proposing that neither side was wholly right nor wrong. Instead, he suggests that violence is not a unitary phenomenon; “some studies address the violence primarily perpetrated by men, while others are getting at the kind of violence that women are involved in as well” (Johnson, 2008, p. 2). Johnson (2008) also called on scholars to empirically test, revise, and refine his proposed typology in order to most effectively understand the differences between types of violence and the role of gender therein. Despite Johnson’s proposed solution being offered over 25 years ago (i.e., that there are two distinct types of IPV), the debate about distinct types of violence and the role of gender persists in IPV research.

Effectively testing the validity of Johnson’s typology and reviewing the research that aims to do so requires understanding the conceptual definitions and related theoretical and methodological assumptions. As such, we provide an overview of the four types of IPV and the core assumptions of the typology.

Violence and Coercive Control

The primary supposition of Johnson’s typology is twofold: there is more than one type of IPV, and the distinctions between types are determined by the context of power and control within which the violence occurs, not the nature of the violence itself (Johnson, 2006, 2008). Because of their centrality to understanding and making distinctions, we start by defining violence and coercive control as they are used in Johnson’s typology. Johnson’s (2008) typology is one of intimate partner violence, meaning that each type includes physically aggressive acts by at least one partner. As such, people in highly coercively controlling relationships absent the use of violence are not included in the typology. Coercive control is distinct from other types of psychological abuse and short-term attempts at control; it is characterized by the persistent use of a variety of nonviolent control tactics intended to exert one’s power over their partner, such as threats and intimidation, surveillance, emotional control and abuse, economic abuse, and isolation (Johnson, 2008; Stark, 2007). Coercive control, especially when accompanied by violence, undermines the will and ability to resist the abusive partner, resulting in entrapment (Stark, 2007). To effectively measure coercive control, researchers must ask about the general use of a variety of control tactics (Johnson, 2008).

Situational couple violence (SCV) is when conflict results in the use of violence by one or both partners and is typically associated with maladaptive communication patterns; it is not rooted in a desire for general control over one’s partner. Intimate terrorism (IT) is when a coercively controlling individual employs violence as part of the strategy to control their partner; IT aligns with popular notions of battering and domestic violence. Violent resistance (VR) is violence enacted by the victim of IT, including violence enacted as an immediate reaction to their coercively controlling partner’s violence and/or violence enacted as retribution. Finally, mutual violent control (MVC) is IPV between two violent, coercively controlling partners (i.e., bidirectional IT). According to Johnson (2008), the frequency and severity of violence have no bearing on IPV classification; IPV types are distinguished solely on patterns of coercive control. To most accurately classify IPV types, information about the use of violence and coercive control from both partners in the relationship is necessary (Johnson, 2008).

Assumptions

Johnson (2008) has identified and tested several core assumptions underlying his typology and reminded scholars, “The task ahead – developing a theoretical framework that recognizes these differences [in types of violence] – will involve the complex scientific process of theory development and empirical testing, followed by theory revision and further testing” (p. 4). Thus, we evaluate the literature testing Johnson’s typology using the following assumptions:

(1) Sampling

Johnson (1995, 2008) asserts that different sampling strategies will largely capture different types of violence. 1a: Clinical and agency samples (e.g., emergency rooms, shelters, agency client lists, law enforcement, courts) are most likely to capture IT and the most serious incidents of SCV (Johnson, 2006, 2008). 1b: General surveys and community samples are most likely to capture SCV. 1c.When participants report on past relationships, representations of IT experiences are more likely to be found in community samples because IT relationships are more likely to end in divorce, and current IT victims are less likely to participate in surveys due to fear of their partner (Johnson, 2008; Johnson et al., 2014).

(2) Gender

In heterosexual relationships, Johnson (2008) asserts that gender is the best predictor of IPV. 2a: Men disproportionately perpetrate IT, and women are more likely to perpetrate VR as the disproportionate IT victims. 2b: Men and women perpetrate SCV at relatively equal rates. 2c: Intimate terrorism is less likely to be bidirectional compared to SCV. 2d: Women are more likely than men to experience fear, other negative psychological outcomes, and injury as the result of IPV, both SCV and IT.

(3) Frequency, severity, and escalation

Although types of IPV are not distinguished based on these characteristics, Johnson nonetheless asserts that differences in frequency, severity, and escalation are likely. 3a: IT is more likely to be frequent, severe, and escalate over time compared to SCV. 3b: Although SCV is less likely to be frequent and severe, there is nonetheless variability in the frequency, severity, chronicity, and outcomes of SCV within and across relationships, including the potential to be lethal.

(4) Antecedents and outcomes

Different types of IPV have different antecedents and outcomes. 4a: IT is rooted in patriarchal norms and motivated by the need to control one’s partner, whereas SCV is the result of poor conflict resolution skills by one or both partners. 4b: IT is more likely to result in injury and other negative physical, and psychological outcomes than SCV.

Method

Our review process was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Page et al., 2021). To ensure the systematic and comprehensive nature of our search, DeSanto, Information and Instruction Librarian, assisted in the development of inclusion criteria, search method and database selection, and the study selection process, and he completed the database searches.

Inclusion Criteria

In line with our study aims, our inclusion criteria were: (a) peer-reviewed, empirical, and original research; (b) direct testing of Johnson’s typology and/or assumptions; (c) published 1995 or later; and (d) studies written in English (including international studies). Qualitative and quantitative studies were included.

Search Method and Study Selection Process

We searched for studies testing Johnson’s typology using the following databases: PsycInfo, PsycArticles, Pubmed, Web of Science, Genderwatch, Sociological Abstracts, and Social Services Abstracts. We searched using terminology from Johnson’s typology of violence since its introduction in 1995, including: intimate terrorism, coercive controlling violence*, situational couple violence*, common couple violence*, violent resistance, mutual violent control, and patriarchal terrorism. We limited initial search results to exclude records published before 1995 and non-English records. The final database search was conducted on March 31, 2021, yielding 881 total records. Conroy and DeSanto reviewed the records for deduplication in Zotero. After deduplication, 502 unique records remained.

Next, Conroy and DeSanto utilized Rayyan for preliminary screening of article titles and abstracts using the inclusion criteria described above. We excluded studies that were unrelated to and/or did not implement Johnson’s typology (e.g., ‘violent resistance’ to political violence); empirical research not published in scholarly, peer-reviewed journals (e.g., conference abstracts, theses and dissertations, etc.); peer-reviewed papers that did not constitute original, empirical research (e.g., commentaries, literature reviews, etc.); and studies that referenced Johnson’s typology as background literature. This yielded 215 studies for full-text review.

Full-text review of articles was used to confirm whether studies provided direct tests of Johnson’s typology and related assumptions. Based on full-text review from Conroy and Crowley, 83 studies tested Johnson’s typology. Next, because conceptual confusion and methodological errors preclude a fair test of Johnson’s typology, Conroy and Crowley each completed another round of full-text review on the 83 studies testing the typology to exclude articles deemed “misapplications” of the typology. Testing a theory (or typology) necessitates accurate representation and pairing with appropriate methods (Tudge et al., 2009); otherwise, “neither refutation nor corroboration is possible” (Tudge et al., 2016, p. 198). Studies were deemed to be misapplications if: the operationalization of coercive control did not align with Johnson’s conceptual definition (see above); if the Negotiation and/or Psychological Aggression subscales of the CTS were used to classify violence victimization/perpetration (given Johnson’s emphasis on physical aggression); IPV types were classified by something other than coercive control (e.g., violence severity or frequency) or classification process was not described; and/or core assumptions were misrepresented (i.e., inaccurate statements about the typology that could be refuted by Johnson’s published work). Of the 83 studies testing Johnson’s typology, we deemed 39 studies to be misapplications to varying degrees (e.g., Bates & Graham-Kevan, 2016; Graham-Kevan & Archer, 2003a; Hines & Douglas, 2010) and, thus, excluded them from analysis.Footnote 3 In total, 44 studies met inclusion criteria for analysis.

Data Extraction and Analysis of Studies

We consulted with one another throughout the screening, full-text review, and data extraction stages. Data extraction was completed in shared online spreadsheets wherein we tracked detailed study information for all full-text reviewed studies and our decision-making processes. We also utilized peer debriefing and discussion when there was disagreement or clarification was needed to determine inclusion or exclusion.

Analysis and Discussion

Summary information from included studies (N = 44) is provided in Table 1, including study aims; sampling and sample characteristics; measures and classifications of violence, coercive control, and IPV type; and findings related to the assumptions of the typology. Of note, not all researchers had the primary aim of testing Johnson’s assumptions as a means of validating his typology. However, by way of testing the typological classifications, studying the effects of sampling, studying the role of gender in IPV, and/or characteristics of IPV (i.e., frequency, severity, and/or escalation), valuable insight is nonetheless provided about the validity of Johnson’s assumptions. We deemed an assumption 'supported’ if there was unanimous support across studies. We deemed an assumption ‘partially supported’ if there were mixed findings across studies and/or if only part of a multipart assumption was supported by study findings. Finally, an assumption was deemed ‘unsupported’ if the existing empirical evidence did not align with the assumption.

Table 1 Summary of included studies, aims, methods, and results (N = 44)
Full size table

Below, we provide additional information about study results, organized by the assumptions of the typology. First, we briefly review the measure of violence and coercive control given the centrality of these two concepts in classifying IPV types. The large majority of quantitative studies included here measured the presence of violence using a version of the Conflict Tactics Scale (CTS), specifically items associated with acts of physical aggression. The number of items varied slightly, from 8–12 items, depending on whether threats of violence and/or sexual violence were included. There was more variety in the measure of coercive control, as seen in Table 1, although researchers largely used previously validated measures. In each case, whether qualitative or quantitative, studies classified types of IPV based on the presence of violence paired with a pattern of coercive controlling behaviors. In most cases, researchers implemented Johnson’s (2008) classification method of “high” control including coercive control scale scores above 2 standard deviations of the sample mean. Although Hardesty et al. (2015) and Eckstein (2016) are particularly useful studies for working toward standardization in the operationalization of coercive control, additional research is needed.

Sampling

We remind readers that Johnson did not suggest clinical and agency samples solely represent IT nor that general surveys only capture SCV. Instead, he argued that SCV is the most common type of IPV in general surveys, while perpetrators and victims of IT were least likely to participate due to fear of consequences (Johnson et al., 2014). Additionally, general surveys about experiences of crime have been associated with higher rates of IT being disclosed (Johnson & Leone, 2005), and reporting on past relationships increases the likelihood of survey participation for IT victims (Johnson et al., 2014). Generally, and as expected, SCV was overwhelmingly represented in the nonclinical samples in our study, and IT was overwhelmingly represented in clinical samples in our study. To test the assumptions regarding sampling, we relied on studies that utilized mixed samples and could directly compare sampling differences.

1a & 1b: Clinical and agency samples are most likely to capture IT and the most serious incidents of SCV, and general surveys are most likely to capture SCV

We explored this assumption using studies that tested differences in type of violence by sample type, i.e., that included comparisons of clinical/agency and general survey samples in a single study (n = 11). Most cases of IT were found in clinical samples, and most cases of SCV were found in nonclinical samples. Similarly, in a meta-analysis of IPV, Love et al. (2020) found that risk markers and characteristics of IT were greater in clinical samples.

1c. When participants report on past relationships, representations of IT experiences are more likely to be found in community samples

Past relationship status was associated with more reports of IT (e.g., Frye et al., 2006; Johnson et al., 2014).

Gender

Although Johnson asserts IT is overwhelmingly perpetrated by men against women in heterosexual relationships, he acknowledges that IT can be perpetrated by women and in same-gender relationships, yet these cases are rarely studied (Johnson, 2008). In our review, only two studies included same-gender relationships (Frankland & Brown, 2014; Hardesty et al., 2008).

Of note, 21 studies included mixed-sex samples, including two studies with dyadic couple data; 21 studies used woman-only samples; and two studies used man-only samples. Studies mostly collected data on victimization alone, whereas 18 studies included data on victimization and perpetration experiences, and one focused only on perpetration. As a result, several studies were limited in their ability to classify relationships beyond SCV and/or IT; only 12 studies also classified VR and/or MVC, thus limiting knowledge about the dyadic nature of IPV.

2a: Men disproportionately perpetrate IT, and women are more likely to perpetrate VR as the disproportionate IT victims

Using mixed-sex samples, 11 studies tested the assumption that men were more likely to perpetrate IT; this was supported by 8 studies, including one study on adolescent dating violence (Messinger et al., 2014). The others found comparable rates of IT for men and women (Brown & Chew, 2018; Jasinski et al., 2014; Lysova et al., 2019). Zweig et al. (2014) found that rates of perpetration varied by respondent sex, with both adolescent girls and boys reporting IT victimization and VR perpetration at higher rates than IT perpetration. Three studies using mixed-sex samples tested and found support for the assumption that women are more likely to use VR than men in response to IT (Graham-Kevan & Archer, 2003b; Messinger et al., 2014).

Moreover, Johnson (2008) and others (e.g., Anderson, 2005; Murnen et al., 2002) suggested that examining rates of violence alone do not capture the nuanced ways that gender affects one’s IPV experience. Some studies in our analysis found this to be the case, with men and women reporting different experiences of IT. For example, male IT perpetrators were more likely to use higher control and more frequent physical aggression than female IT perpetrators (Graham-Kevan & Archer, 2003b, 2008; Michalski, 2005). However, others found that men and women IT perpetrators did not differ on the frequency or severity of violence (Jasinski et al., 2014); and male IT victims reported more severe victimization than female IT victims (Lysova et al., 2019).

2b: Men and women perpetrate SCV at relatively equal rates

Using mixed-sex samples, 9 studies tested this assumption, with six finding support, including in one adolescent sample (Gaertner & Foshee, 1999). The remaining three studies found that men were more likely to report SCV victimization; similarly, Graham-Kevan and Archer (2008) found that female SCV perpetrators used more acts of physical violence. Zweig et al. (2014) again found that rates of perpetration varied by respondent sex, with girls reporting more gender symmetry in their SCV experiences and boys reporting more unilateral SCV from their female partners. Of note, studies that gather only victimization data are unable to account for bidirectional SCV.

2c: Intimate terrorism is less likely to be bidirectional compared to SCV

Only three studies tested this assumption, with two finding a higher rate of bidirectional SCV than MVC. On the other hand, Messinger et al. (2014) found that most adolescent SCV was unilateral, whereas adolescent MVC (bidirectional IT) was more common than unilateral IT.

2d: Women are more likely than men to experience fear, other negative psychological outcomes, and injury as the result of IPV, both SCV and IT

Using mixed-sex samples, six studies tested and provided at least partial support for this assumption. More specifically, Johnson et al. (2014) found that male IT perpetrators were more likely than female IT perpetrators to make their partner feel inadequate and fearful, and to economically entrap their partner. Moreover, regardless of IPV type, women were more likely to be injured, depressed, and use painkillers than male victims. Others found that men and women respondents reported male IPV perpetrators as inflicting more injury (Brown & Chew, 2018; Graham-Kevan & Archer, 2008; Jasinski et al., 2014; Lysova et al., 2019). Eckstein (2016) found that women IT victims experienced more IPV-related stigma than other victims, and Jasinski et al. (2014) found that female IT victims were more likely to leave the relationship than men. Contrary to Johnson’s assumption, however, Jasinski et al. (2014) found no gender differences between IT victims on measures of PTSD, depression, nor missed paid work. Similarly, Lysova et al. (2019) found no gender differences in the long-term effects (i.e., PTSD) related to IPV. Additionally, Graham-Kevan and Archer (2008) found no significant gender differences in injury from SCV.

Providing rare insight into the nature of MVC as compared to other IPV types, albeit based only on women’s reports, Graham-Kevan et al. (2012) found that male IT perpetrators used less injurious violence compared to that of men in MVC relationships and men using VR.

Frequency, Severity, and Escalation

Johnson’s classifications of IPV are not based on the frequency, severity, or escalation of violence; rather, he argued that although IT was more likely to be frequent, severe, and escalate over time, SCV could be chronic and/or lethal (Johnson, 2008).

3a: IT is more likely to be frequent, severe, and escalate over time compared to SCV

This was the second most tested assumption (n = 27), particularly regarding frequency and severity. As noted in Table 1, there was overwhelming support for this assumption, with only two exceptions. Hardesty et al. (2016) found no significant difference in violence frequency nor severity between IT and SCV groups. Also contrary to prediction, Frankland and Brown (2014) found higher rates of physical aggression in SCV compared to IT. Jasinski et al. (2014) found partial support for the assumption, with male IT victims reporting more frequent, but not more severe, violence compared to SCV.

In addition to finding more frequent and severe physical violence in IT compared to SCV, some studies also found that IT perpetrators were more likely to use sexual coercion and violence (Frankland & Brown, 2014; Graham-Kevan et al., 2012; Hines & Douglas, 2018; McKay et al., 2020; Verschuere et al., 2021). Young adults exposed to marital IPV as children and adolescents also reported being exposed to more frequent and severe physical violence (Haselschwerdt et al., 2019, 2020) and nonviolent abuse tactics in IT parental relationships, including being used as a tool of abuse by their abusive fathers (Haselschwerdt et al., 2021).

A small number of studies also provided rare insight into the nature of violence in MVC relationships. For example, studies found that MVC relationships were characterized by higher rates of physical and/or sexual violence compared to other IPV types (Frankland & Brown, 2014; Graham-Kevan et al., 2012; Messinger et al., 2014). Frankland and Brown (2014) also note that the unexpectedly high rate of MVC in their sample of same-sex couples relative to MVC rates found among heterosexual couples suggest “that gender is a crucial factor” to understanding manifestations of IT (p. 21). Zweig et al. (2014) found that compared to SCV, adolescent IT perpetrators experienced more frequent and severe violence by their partners using VR.

3b: Although SCV is less likely to be frequent and severe, there is nonetheless variability in the frequency, severity, chronicity, and outcomes of SCV within and across relationships

Few studies (n = 6) explore the variability of SCV; however, as noted in Table 1, all six studies indicate that SCV can be very frequent, very severe, and/or escalate over time. Moreover, Nielsen et al. (2016) found that some female SCV victims were fearful of their partners during marriage and post-separation, at which time they implemented protective strategies for themselves and reported IPV-related depression and PTSD.

Antecedents & Outcomes

4a: IT is rooted in patriarchal norms and motivated by the need to control one’s partner, whereas SCV is the result of poor conflict resolution skills by one or both partners

Compared to other assumptions, there was greater variability of methods testing this assumption regarding the motives underlying IT and SCV. However, there was consistent support for this assumption among the 12 studies to do so. In a study of cultural attitudes about gender equity, Brownridge (2010) found that women in a less patriarchal region of Canada experienced lower levels of IT compared to more patriarchal regions. This finding suggests that greater cultural endorsement of gender equity may influence attitudes toward and use of violence against women.

At the relationship level, couple dynamics also differed in stark ways between IT and SCV, with IT being associated with significantly higher levels of psychological aggression and unique patterns of control. For example, among divorcing mothers who experienced IPV, IT victims experienced more post-separation harassment from ex-partners (Hardesty et al., 2015, 2017). Qualitative reports of women’s IPV experiences also suggested that IT victims were more likely to describe their experiences as serious and life threatening, with violence being part of a pattern of power and control (Tiwari et al., 2015). IT was also characterized by notably greater use of control tactics such as destroying property and monitoring (Hardesty et al., 2008). In a qualitative study of incarcerated, male IT perpetrators, McKay et al. (2020) found that IT perpetrators lacked empathy; planned to use the legal system against their partners, including to take away the children; tried to discredit and blame their partner for the violence; and did not take responsibility for their own violent behaviors. Hardesty et al., (2015, 2016) empirically confirmed that differences in outcomes between IT and SCV victims are associated with the control context, not the physical violence, thus highlighting the insidiousness of coercive control as a characteristic of IT.

On the other hand, SCV was not found to be rooted in ongoing partner efforts to control the victim. For example, women’s qualitative reports of SCV described violent relationships as being characterized by recurring arguments (Tiwari et al., 2015). This aligns with young adult reports of the IPV that they were exposed to as children and adolescents: Conflicts led to physical violence and, in some cases, violence ceased when stressors and relationship challenges were resolved (Haselschwerdt et al., 2019, 2020). Similarly, Stith et al. (2011) found that SCV was rooted in vulnerabilities and stressors and tied to specific arguments; respondents also seemed to perceive the use of violence as an adaptation rather than inherent to their relationships.

4b: IT is more likely to result in injury and other negative physical, and psychological outcomes than SCV

This was the most commonly tested assumption (n = 29). As noted in Table 1, all 29 studies provided at least partial support for this assumption using quantitative and qualitative methods. We focus our discussion here on the less often studied outcomes, beyond injury, PTSD, depression, prescription drug use, and fear, except in cases where support did not hold up for one of these more commonly studied outcome variables. Eckstein (2016) found unique psychosocial outcomes for IT victims, who reported more IPV-related stigma than SCV victims and were more likely to use situational withdrawal to cope. Similarly, others found that IT victims perceived having less social support (Leone et al., 2007) and experienced more suicidal behavior (Leone, 2011). Leone et al. (2004) also found that IT victims were more likely to report needing help and to miss more work and other activities.

With regard to relational outcomes, Eckstein (2012) found that IT victims also experienced greater partner and relational uncertainty. Moreover, Johnson and Leone (2005) found IT victims were more likely to leave and do so a greater number of times than SCV victims, as well as more likely to seek out safe locations for residence. Hardesty et al. and’s (2015, 2016) work also provides valuable insight into couple dynamics following divorce, with IT victims reporting more frequent post-separation harassment, perceived threat of future harm, and fear of partner after separation. Following separation, IT victims also placed less importance on the maintenance of the father-child relationship (Hardesty et al., 2016), and greater post-separation conflict, less communication about child rearing, and lower levels of support compared to SCV victims (Hardesty et al., 2017).

Moreover, Haselschwerdt et al., (2019, 2021) expanded the focus of Johnson’s typology to the family unit in their studies of young adult exposure to marital violence during childhood and adolescence. Like IT victims, children exposed to parental IT experienced notably poorer outcomes. For example, children exposed to IT were subjected to more nonphysical abuse tactics, as well as more frequent and severe physical violence than those exposed to SCV (Haselschwerdt et al., 2021). IT exposure also corresponded to their own direct experiences of abuse by their fathers, including more harsh parenting and more ongoing child abuse (Haselschwerdt et al., 2019).

Once again, few studies provide insight into MVC and VR. However, studies suggested that MVC is associated with the greatest range of violent tactics and injury among all violence types (Frankland & Brown, 2014; Hines & Douglas, 2018), and that MVC men are more injurious than male IT perpetrators (Graham-Kevan et al., 2012). Perhaps more surprisingly, IT perpetrators were found to be less injurious than men using VR (Graham-Kevan et al., 2012).

Methodological Limitations in the Literature and Recommendations for Future Research

Although we implemented inclusion criteria to include studies with sound conceptualization and operationalization of Johnson’s typology, our review of included studies nonetheless suggests that even accurate, high-quality research on Johnson’s typology has limitations to be considered now and in future research. First, several studies measured violence victimization (and in one case, perpetration), rather than reports of victimization and perpetration. As a result, missing dyadic information precludes the ability to most accurately and comprehensively classify IPV types, and to best understanding the complex nature of IPV. Similarly, single-sex samples make it difficult to meaningfully compare men and women’s experiences of perpetration and victimization. As such, future research would benefit from mixed-sex samples with respondents reporting on victimization and perpetration so that the typology and related assumptions can be more comprehensively tested. Similarly, we note (and as seen in Table 1) that the majority of the literature does not often test multiple assumptions in a single study; instead, Johnson’s typology is used as part of a study with broader aims than testing typology itself. Additional research designed with the primary purpose of testing Johnson’s typology is needed to further validate, revise, and expand on the current classifications of SCV, IT, VR, and MVC.

One additional limitation of the literature is inherent to an arguable limit of Johnson’s typology, which is the omission of nonviolent coercive control. To be fair, the typology was developed for studying violent relationships; as a result, nonviolent coercive control is not captured by Johnson. Nevertheless, he (2008) and others (Stark, 2007) acknowledge the debilitating nature of coercive control, even in the absence of violence. However, few studies focus on coercive control in the absence of violence. Failure to study nonviolent coercive control in its own right precludes better understanding of the relationship contexts that may escalate to include physical violence. Fortunately, emerging research provides insights about dynamics of coercive control and physical violence not originally considered by Johnson, including nonviolent coercive control. For example, Anderson (2008) found that high coercive control is associated with negative health outcomes even in the absence of physical violence. Similarly, Crossman et al. (2016) found that nonviolent coercive control victims reported more fear post-separation than IT mothers. However, others have found that using a continuous measure of violence (i.e., total number of violent acts) may be a better predictor of some outcomes, such as likelihood of injury, leaving the relationship, or depression (Anderson, 2008; Bubriski-McKenzie & Jasinski, 2013) as compared to the predictive ability of typological differences (i.e., IT or SCV).

Others have studied additional combinations of violence and coercive control, including the use of violence in response to a nonviolent coercive controling partner, further expanding Johnson’s typology (Conroy & Crowley, 2021; Zweig et al., 2014). Brown and Chew (2018) also suggested that MVC may actually include the use of violence and coercive control as a counter-control tactic by one partner. Together, these findings suggest that Johnson’s (2008) brief acknowledgement of “incipient intimate terrorism” (p. 46), or nonviolent coercive control, deserves greater attention in the IPV research, in its own right, and for its potential to escalate to IT. To do this, data about violence victimization and perpetration are needed, as well as information on violence frequency and severity; use of coercive control; respondent and partner gender; and motives for and outcomes of violence. As Johnson (2008) notes, capturing these nuances may be best achieved using qualitative and/or mixed methods, with mixed methods providing the added benefit of triangulation. For example, McKay et al. (2020) found that quantitative classifications of MVC were challenged by qualitative reports where participants did not perceive their partner as controlling, despite relatively high jealousy-related control items; instead, their qualitative narratives were more fitting of IT or SCV and, thus, provided more nuance than quantitative measures alone.

Additionally, longitudinal research designs are rarely implemented (exceptions include Hardesty et al., 2017, 2019; Haselschwerdt et al., 2016) but can provide useful information about the changing nature of physical violence and coercive control over time. As Leone et al. (2004) found, there was a unique type of coercive control that emerged, characterized by no threats; the authors suggested that IT may reduce the need for threats over time because, with violence, the control becomes more credible. Similarly, studies testing Johnson’s typology reviewed herein overwhelmingly rely on convenience sampling (exceptions are noted in Table 1 when meta-analyses, national representative samples, or random sampling strategies were used), and data are largely gathered in the United States [exceptions include studies from Singapore (Brown & Chew, 2018), Canada (Brownridge, 2010; Lysova et al., 2019; Michalski, 2005), Australia (Frankland & Brown, 2014), England (Graham-Kevan & Archer, 2003b, 2008), Mozambique (Graham-Kevan et al., 2012), China (Tiwari et al., 2015), the Netherlands (Verschuere et al., 2021), and Love et al.’s (2020) meta-analysis of U.S. and international studies]. Future research would benefit from testing the validity of Johnson’s typology across cultural contexts.

Moreover, the assumptions underlying Johnson’s typology are derived from research on heterosexual couples; thus, the literature testing the typology overwhelmingly reflects the experiences of cisgender individuals and heterosexual couples. However, if gender is instead conceptualized as individual, relational, structural, and cultural (Anderson, 2005) – and is subsequently integrated into methods as such – researchers can move beyond debating the role of gender to exploring it in all experiences of IPV, including the unique relevance of gender to men’s experiences of IPV, as well as that of same-sex couples, queer couples, and transgender individuals. For example, research studies can ask questions about individual-level variables (e.g., gender ideologies), relationship-level variables (e.g., gender equity in the relationship), structural-level variables (e.g., experiences with formal help-seeking), and cultural-level variables (e.g., national gender equity) to better understand how gender affects experiences of IPV beyond individual identities. In other words, the research question of interest that will produce more fruitful findings is not ‘are there gender differences in experiences of IPV?’ but ‘how does gender influence experiences of IPV?’ This also allows for testing the assumption that violence is rooted in gender without being perceived as mutually exclusive from, or at odds with, acknowledging men’s experiences of IPV and the role of gender therein, nor the relevance of other factors to IPV. In research, there is room for examining all relevant factors at play in IPV, and Johnson’s typology provides but one lens to implement.

Similarly, notions of power and control reflected in the literature are tied closely to early iterations of The Duluth Model, developed for use with heterosexual couples; this narrow focus is reflected in the notably smaller volume of research on same-sex couples, genderqueer couples, and transgender individuals. To more effectively understand the ways that gender uniquely affects queer couples and individuals, and the variety of other ways (in addition to gender) that power and control can manifest for systemically marginalized individuals, it is necessary to include same-sex couples and transgender individuals in research on IPV. Doing this effectively would necessitate using measures of coercive control that capture abuse tactics as well as theoretical frameworks that are representative of the unique experiences of the population being studied. Such IPV research exists and provides useful frameworks for more critical conceptualizations of power and control that can be applied to testing Johnson’s typology (e.g., Dolan & Conroy, 2021; Bermea et al., 2021; Linder, 2018). Correspondingly, there is limited use of Johnson’s typology with samples of racially and ethnically marginalized individuals (exceptions: Bubriski-McKenzie & Jasinski, 2013; Graham-Kevan et al., 2012; Leone, 2011; Leone et al., 2004, 2007; Tiwari et al., 2015). To address this limitation, more intersectional feminist perspectives of gender would be particularly useful.

Overall, in order to meaningfully interpret findings within and between studies related to Johnson’s typology, it is essential for researchers to transparently and clearly report on their methods, including clearly stated hypotheses, sampling strategies, measures of coercive control and violence, and analytic techniques for classifying types of violence and hypotheses testing. To do this, researchers are encouraged to look to Johnson’s (2008) conceptual definitions and recommendations for measurement. Moreover, several studies provide useful information on the validity and utility of coercive control measures and classification techniques (e.g., Eckstein, 2017; Hardesty et al., 2015; Lehmann et al., 2012).

Study Limitations and Contributions

As a result of our inclusion criteria, this study does not fully capture the debate regarding gender and IPV, the scope of research suggesting that IPV is a unitary phenomenon, nor other typologies proposed in the IPV literature. Our review was also limited to empirical, peer-reviewed research, which excludes other published IPV literature. We chose to include only peer-reviewed, empirical research because of the necessity for hypothesis testing in theory building, and, in this case, testing Johnson’s assumptions about types of IPV. In doing so, the widespread scholarly debate about the role of gender in IPV that motivated this study is not represented herein. Of note, however, our original analysis included data extraction and assumption testing for all 83 articles testing Johnson’s typology, including misapplications. This resulted in an overwhelming volume of “mixed results,” with Johnson’s typology both widely supported and challenged. Making sense of these findings proved difficult without extensive critique of the methods used, which did not allow for us to focus on our primary research question about the state of Johnson’s typology since its inception. However, upon the suggestion of reviewers and given the necessity of accurate conceptualization and operationalization to fairly refute or corroborate Johnson’s assumptions, we ultimately removed the misapplications and alternative uses (e.g., brief mention of typology as background literature without empirical testing of the typology) of Johnson’s typology from our analysis. By including only studies that correctly represented and operationalized the typology, the findings from included studies presented herein resulted in overwhelming support for Johnson’s typology, with only few exceptions, as noted in the analysis and discussion section. This is a particularly important finding as it speaks to the importance of accurate representation and operationalization of a given typology, model, or theory for achieving the desired outcome of testing that typology, model, or theory. This is the first systematic review of Johnson’s typology in the IPV literature. Despite the ongoing debate about the role of gender and IPV that persists today, findings from the studies spanning 22 years of testing Johnson’s typology presented herein provide overwhelming support for the argument that there are distinct types of violence, and gender is central to understanding those distinctions.

Conclusion

We systematically reviewed the peer-reviewed, empirical literature testing Johnson’s (2008) typology of violence in order to evaluate the progress made in testing and revising Johnson’s typology by reviewing, summarizing, and critically evaluating the available literature. In total, 44 articles meeting our inclusion criteria tested one or more assumptions of Johnson’s typology and provided overwhelming support for the notion that there are several distinct types of violence, distinguished by patterns of coercive control. With few exceptions, research supports the notions that clinical samples are more likely to capture IT and nonclinical samples are more likely to capture SCV; that IT is most often perpetrated by men against women, is more likely to be frequent and severe, and result in negative outcomes than SCV; and that IT is rooted in patriarchal norms and control motives, whereas SCV is situated in particular conflicts.