Intimate Partner Violence Perpetration Denial and Underreporting in Cisgender Male Couples

Introduction

Although IPV is often heteronormatively-framed and gendered, with female victims/survivors and male perpetrators, the prevalence of IPV in sexual and gender minority (SGM) partnerships may be as high, or higher, than among non-SGM couples (Finneran & Stephenson, 2013a; M. Liu et al., 2021; Rollè et al., 2018). Indeed, research suggests that the additional minority stress faced by SGM increases IPV risk (Callan et al., 2021; Edwards & Sylaska, 2013; Stephenson & Finneran, 2017). Yet, IPV-mitigation in these vulnerable groups is hampered by a combination of individual- and systemic-factors (Scheer et al., 2020), including minimization and perceptions that abuse among SGM is less severe or problematic (Alhusen et al. , 2010; Finneran & Stephenson, 2013a; Murphy-Oikonen & Egan, 2022; Poorman et al., 2003); identity-related stigma (Calton et al., 2016; M. Liu et al., 2021); and a paucity of relevant research, services, and resources (Calton et al., 2016; Edwards et al., 2020; Kim & Schmuhl, 2021). Notably, there are substantial gaps in our understanding of IPV perpetration and perpetrators in SGM, including within cisgender-male couples (Graham et al., 2021; Thompson et al., 2006). Understanding the individual, relationship, and contextual factors that precede, prompt, and sustain perpetrators’ abusive behavior is critical to reducing IPV incidence and prevalence (M. Liu et al., 2021). Therefore, it is imperative that we can accurately identify perpetrators, characterize their commonalities, and design targeted prevention and reduction programs (Sheehan et al., 2012). However, identifying and targeting perpetrators for research and intervention is limited by multiple factors, including offenders’ underreporting and denial (LaMotte et al., 2014).

Given the potential legal and social repercussions of admitting to abusive behavior, denial among IPV perpetrators is unsurprising. Violent-, including sexual-, offender denial has been repeatedly noted in criminal and clinical populations, including samples drawn from court-mandated programs/interventions and incarcerated sexual offenders (Barbaro & Raghavan, 2018; Dietz, 2020; Heckert & Gondolf, 2000; Smith, 2007). In these populations, denial is a well-documented strategy that ranges from outright repudiation (often referred to as “categorical denial”) to minimization (Scott & Straus, 2007), which can include admitting to abusive acts but reporting lower severity than evidence suggests, shifting blame to external forces or their partner, and/or suggesting that their actions were misinterpreted (Dietz, 2020; Morrison et al., 2021). Often, perpetrators deny their offenses to avoid negative consequences, such as arrest, and/or facilitate positive outcomes, such as being released from mandated programs or during parole review (Henning & Holdford, 2006). In rehabilitation and punitive settings, denial can have significant implications for IPV prevention and intervention efforts, as both barrier to treatment and/or an indicator that an individual is not ready to change (Morrison et al., 2021). Research has associated denying and minimizing domestic abuse with more severe abuse and aggression (Heckert & Gondolf, 2000; Helfritz et al., 2006); repeat offenses and recidivism (Barbaro & Raghavan, 2018; Senkans et al., 2020); and lower treatment compliance (Henning & Holdford, 2006; Kropp & Gibas, 2020; Morrison et al., 2021; Scott & Straus, 2007). Moreover, although helping abusers accept culpability for their behavior is often a key component of interventions (Scott & Straus, 2007; Sheehan et al., 2012), for some offenses categorical denial is used as an exclusion criteria for treatment (Ware et al., 2020). Therefore, reducing denial and increasing individual accountability is critical to effectively rehabilitating identified offenders, and reducing future harm to partners.

Outside of legal and criminal consequence-avoidance, research suggests that social desirability bias is a primary driver of IPV denial, as it represents a perpetrator’s attempt at image management (Bell & Naugle, 2007; Follingstad & Rogers, 2013; Freeman et al., 2015). Indeed, studies have noted that many incarcerated perpetrators continue to deny their actions after conviction and in the face of substantial physical evidence (Bourke et al., 2015; Ware & Blagden, 2020). This suggests that conscious deception is not the only source of offender-misinformation; denial may also be an unconscious strategy that protects self-image by shifting accountability away from the individual (Schneider & Wright, 2004). In addition, there is evidence that memory and recall can contribute to perpetrators’ denial and underreporting (Halim et al., 2018; Medina et al., 2004). Given its underlying complexities, denial is, therefore, in-and-of itself, a valuable research target, and could provide valuable information about conscious and unconscious cognitive biases that could be leveraged in IPV research and reduction programs (Dietz, 2020). However, as many of these results come from studies conducted in populations that skew towards more severe, violent abusers, inference from this research may not be generalizable to less severe and/or non-physically injurious types of IPV.

Perpetration denial has also been documented in survey-based IPV research, where data is generally self-reported, without external validation or corroborating information (e.g., criminal complaints, observed injuries). From a research standpoint, valid IPV data is critical to addressing IPV on a broad scale and accurately estimating IPV burdens. Yet, this research area is overwhelmingly cross-sectional and reliant on survivor’s self-reported experiences; crucial information on perpetration is therefore often second-hand, and thus may not capture key cognitive, psychosocial, and contextual factors antecedent to abuse (Kim & Schmuhl, 2021; M. Liu et al., 2021). In addition, IPV prevalence estimates based on self-reported perpetration tend to be lower than those based on self-reported victimization, both within and across study populations (Armstrong et al., 2002). Dyadic concordance research, which compares partners’ corresponding perpetration-victimization reports to assess data reliability, has repeatedly documented partners’ incompatible IPV reports across diverse study samples, data granularities, and IPV types (e.g., Cui et al., 2005; Cunradi et al., 2009; Kuijpers, 2020; LaMotte et al., 2014; Marshall et al., 2021; Walsh & Stephenson, 2022; Wenger, 2015). This area of research confirms that perpetration-underreporting, or denial, is not limited to criminal populations or the most serious offenders. Due to this underreporting and denial, misclassification bias may be a substantial issue in IPV research (Ryan, 2013; Schafer et al., 2002). Misclassifying perpetrators as non-offenders in analyses may bias results and lead to potentially invalid inference, further hamstringing our collective ability to understand and address IPV holistically.

Thus, there are significant barriers to investigating and addressing IPV perpetration. Moreover, IPV research, both of population-based and criminal samples, has historically consisted of investigations into male aggression and perpetration against female victims, often in the context of mixed-gender couples. As a result, despite the additional risk factors and high IPV prevalence borne by SGM, our understanding of IPV in populations of gay, bisexual, and other men who have sex with men (GBMSM) is minimal (Finneran & Stephenson, 2013b; Kim & Schmuhl, 2021; M. Liu et al., 2021). Nascent research does, however, suggest that SGM perpetrator typologies are complex, heterogeneous, distinct from cisgender, heterosexual typologies, and may not be reliably captured with traditional IPV survey instruments (Donovan & Barnes, 2020). Additionally, although there is evidence that GBMSM perpetrators may be less prone to underreport and deny their behavior than cisgender-heterosexual perpetrators (Stephenson et al., 2019; Walsh & Stephenson, 2022; Wu et al., 2015), current SGM-IPV prevalence estimates are wide-ranging enough to suggest systematic measurement bias. Thus, despite increased awareness of IPV in SGM relationships and the need for expanding SGM-inclusive IPV and perpetrator research, significant questions about perpetration reporting, denial, and measurement remain.

The goal of this secondary analysis was to investigate the prevalence and correlates of IPV denial, defined here as underreporting, across three types of abusive behavior—emotional, monitoring/controlling, and physical/sexual. Specifically, we assessed interpartner agreement and patterns in discrepant self-reported perpetration and partner-reported victimization across IPV types in a convenience sample of cisgender GBMSM-couples. In addition, we analyzed individual-, partner-, and couple-level correlates of denying emotional, monitoring/controlling, and physical/sexual perpetration, including individual- and partner-reported SGM-related experienced stigma. This study extends the handful of previous studies assessing dyadic concordance in IPV reports from male partners (Stephenson et al., 2019; Walsh & Stephenson, 2022; Wu et al., 2015), and the first of which we are aware to investigate perpetration denial in GBMSM.

Method

Population

Data for these analyses was taken from baseline survey data collected between April 2016 and June 2017 as a part of a randomized control trial of video-based couples HIV counselling and home-based testing (CHCT) in the United States (ClinicalTrials.gov Identifier: NCT02335138; detailed methods have previously been described; Stephenson, Freeland, et al., 2017). Male couples were recruited via advertising on social media websites and mobile apps, and were eligible to participate if both partners were (a) > 18 years, (b) identified as cisgender male, (c) in a sexual relationship with each other for > 6 months, (d) had not had an HIV test in the past 3 months, (e) did not report severe intimate partner violence (IPV) in their relationship within the past year (severe IPV was defined as non-consensual punching, hitting, slapping, kicking, rape, and/or forced sexual activity, as perpetrator or victim/survivor), and feeling unsafe in their relationship), (f) were willing to receive rapid home HIV test kits, (g) had internet access, (h) self-reported concordant HIV-negative or HIV-serodiscordant, and (i) did not report coercion to participate in HIV testing or the study. Upon consent, each partner separately completed baseline surveys, reporting sociodemographic, psychosocial, and relationship characteristics. The study was reviewed and approved by the University of Michigan Institutional Review Board and a Data Safety Monitoring Board.

Eight-hundred and fifty-seven men completed the baseline survey. For the current study, participants without corresponding partner-reported data (n = 7) and those missing all self- and partner-reported IPV data (n = 2) were excluded, resulting in 848 individuals (98.95%), or 424 male couples, in the analytic sample.

IPV Measurement

Past-year IPV victimization and perpetration were measured with the Gay and Bisexual Men (IPV-GBM) scale, a validated scale developed for use in GBMSM (victimization experience: Cronbach’s alpha > .78; perpetration: Cronbach’s alpha > .76) (Stephenson & Finneran, 2013). The IPV-GBM scale measures 13 corresponding victimization experiences and perpetration behaviors across emotional, monitoring and controlling, and physical and sexual domains. This scale has previously been used to measure IPV prevalence in GBMSM populations (Stephenson et al., 2019; Stephenson, Suarez, et al., 2017). Participants first reported how often they had experienced each abusive behavior with their primary partner in the previous year, and then how often they had perpetrated each of the same behaviors against their partner. Respondents were instructed to exclude any consensual acts, such as mutually-agreed upon BDSM, from their reports. Frequencies were chosen from the following: never, once, twice, 3-5x, 6-10x, 11-20x, > 20, not in past year but before, don’t know. For the current study, individual-level IPV variables were constructed by aggregating individual responses according to IPV (overall and by domain), and dichotomized as: (a) not [IPV/domain] [victimization/perpetration] in the past year or (b) [IPV/domain] [victimization/perpetration] at least once in the past year; “don’t know” was treated as missing data. We followed RAINN’s recommended language usage in the current study—using “victim” to describe recent violence and/or when discussing a particular crime, as opposed to “survivor,” which more generally refers to someone who has gone through the recovery process (Rape Abuse & Incest National Network [RAINN, 2023]).

Perpetration and Perpetration Denial

For each domain, individual participant’s perpetration report was compared to their partner’s victimization report to assess perpetration denial. [Domain] perpetrators (e.g., emotional perpetrators) were defined as individuals who reported perpetrating at least one behavior within a given domain and/or whose partner reported one or more experiences within a given IPV domain. These perpetrators were then dichotomized according to reporting-concordance; those who reported “not perpetrating” IPV but whose partner reported victimization (within a given domain) were considered perpetration deniers, in contrast to those with concordant perpetration and victimization reports. Note that partners were not presented with each other’s IPV reports, so individual self-reporting and denial was not based on reactions or responses to a partner’s reported experience(s).

Independent Variables

Individual, partner, and couple characteristics that have been associated with IPV perpetration and/or perpetration reporting in GBMSM populations were analyzed as independent variables (race/ethnicity, sexual orientation, education, employment, substance use, depression, and experienced homosexual stigma). Race and ethnicity were combined into a single variable, as multivariate models did not converge without combination. Participants reported their sexual orientation as either gay/homosexual, heterosexual, bisexual, queer, questioning, or other—this was dichotomized as gay and bisexual, queer, or questioning, based on the small proportion of individuals who did not self-identify as gay/homosexual (9.08%). Dichotomized substance use was constructed according to self-reported binge drinking (> 5 drinks on at least one occasion in the previous 3 months) and/or recreational drug use (any use in the past 3 months). Experienced homosexual stress/stigma was measured with the Heterosexist Harassment, Rejection, and Discrimination Scale (HHRDS) scale (Szymanski, 2006) and depression was assessed using the CESD-11 Iowa form (Kohout et al., 1993). At the couple-level, marital status and relationship length were assessed, and dyadic differences in age, experienced stigma, and depression were constructed by subtracting partner-values from self-reported data.

Imputation

Missingness was assessed for all dependent and independent variables. There was complete data for all 3 dependent, perpetration denial variables. Approximately 86% of the full study sample had complete independent variable data (N = 728); 118 individuals were missing data for 1 variable (experienced stigma: n = 92 [10.85%]; depression: n = 23 [2.71%]; substance use: n = 4 [< 1%]; employment: n = 1 [< 1%]); and 1 participant was missing substance use and experienced stigma data. Multiple imputation (MI) was used to replace missing values in predictor variables using the distribution of complete observed and missing values with SAS MI procedures (fully conditional specification [FCS]; 20 imputations), using appropriate imputation models (e.g., logistic regression for binary variables). There were no discernible patterns in missingness across variables, so MI procedures were performed under the assumption of missing at random (MAR). In addition to model variables, internalized homophobia (Smolenski et al., 2010), anticipated homosexual stigma (H. Liu et al., 2009), and HIV status (binary) were used as auxiliary variables in MI. Scale-based continuous variables were constrained according to the variable’s possible value range. The full sample of 848 participants was used for the MI procedure. Diagnostic trace plots indicated convergence for each imputed variable, i.e., that the chains reached their appropriate stationary posterior distributions. All reported results are based on imputed data.

Tabla 1

Individual and Relationship Characteristics, and Domain-Specific Intimate Partner Violence (IPV) erpetration and Victimization in a Sample of Cisgender Male Couples (n = 848 Individuals), and a Subsample (n = 663) Containing Self- and/or Partner-Reported IPV Perpetrators (2016-2017, United States)

Note. ¹Men who self-reported perpetration at least one form of IPV in the previous year and/or whose partner reported at least one IPV victimization experience in the previous year. ²Other race includes: native American or Alaskan native; native Hawaiian or other Pacific islander. ³Past 3 months; recreational drug use refers to any self-reported amphetamine, methamphetamine, hallucinogen, club drug, cannabis, analgesic narcotic, or other illicit substance use. ⁴Heterosexist Harassment, Rejection, and Discrimination Scale (HHRDS); possible values 18-84, higher values indicated higher experienced stigma. ⁵CESD-11 Iowa form; possible values 0-22, higher values indicate higher depressive symptoms. ⁶Includes 298 dyads where both partners were categorized as perpetrators and 67 individuals whose partners were not categorized as perpetrators. ⁷Absolute difference between male partners.

Tabla 2

Patterns of IPV Perpetration Reporting and Denial Across Emotional, Monitoring/Controlling, And Physical/Sexual IPV Domains, in a Sample of IPV Perpetrators in Cisgender Male Couples (N = 663; United States, 2016-2017)

Analysis

Appropriate distributions of individual, partner, and relationship characteristics were assessed for the full study sample (N = 848) and the sub-sample of IPV perpetrators (n = 663). For each IPV domain (emotional, monitoring/controlling, physical/sexual), distributions of individual, partner, and dyadic variables were compared between self-reported perpetrators and perpetration deniers, using two-sided Kruskal-Wallis tests for continuous variables and chi-square (or Fisher’s exact) tests for categorical. Each domain-specific bivariate analysis was restricted to individuals categorized as [domain] perpetrators (emotional: n = 527; monitoring/controlling: n = 490; physical/sexual: n = 267). These descriptive analyses did not account for interdependence between partners.

We assessed the associations between actor-, partner-, and dyadic-level effects and perpetration denial for each of the three IPV domains using actor-partner interdependence models (APIM), using recommended generalizing estimating equation-methods for binary outcomes (Loeys & Molenberghs, 2013). Crude models were fitted to assess the odds of denying a given type of IPV perpetration and each of the independent variables. Moderation of actor race, education, employment, substance use by partner variables was assessed via statistical significance of the interaction term, as per Aiken et al., 1991; no significant moderation was identified for any of the three IPV outcomes and no interaction terms were included in the full models. Adjusted models contained all actor and relationship characteristics, as well as categorical partner measures. Individual- and partner-age were centered at 18, the study’s minimum age. To avoid structural multicollinearity issues, for continuous measures (age, experienced stigma, and depression), either the partner-measure or the difference between the actor and partner variables were included as independent factors, depending on which variable had the strongest crude association with a given type of perpetration denial.

Results

The study population’s sociodemographics and IPV reports are presented in Table 1. The majority of the sample was non-Hispanic White, employed, had a college degree, self-identified as gay, and reported binge drinking and/or using recreational drugs in the previous 3 months. Over 80% of the study’s couples had been together for at least 1 year and slightly more than two-thirds of the study’s couples were married. Forty-five percent, 45.40%, and 20.05% of the sample were categorized as emotional, monitoring/controlling, and physical/sexual IPV perpetrators, respectively. Of the 848 analyzed individuals, 185 (21.80%) were not categorized as IPV perpetrators of any kind, i.e., they neither self-reported perpetration nor did their partner report victimization. Within the 663 identified perpetrators, sociodemographic characteristics were similarly distributed to the full study sample.

Table 2 presents the perpetration-reporting and denial patterns and overlap, by IPV domain. Emotional IPV perpetration had the highest prevalence, representing 79.49% (n = 527; 101 individuals and 214 dyads) of the sample, closely followed by monitoring/controlling (n = 490 (73.91%); 110 individuals and 190 dyads). Physical/sexual IPV perpetration was the least common type in the study (n = 267 (40.27%); 87 individuals and 90 dyads). Conversely, physical/sexual IPV perpetration was denied to a greater extent than the other types; 36.33% (n = 97) of identified physical/sexual perpetrators did not self-report their behavior. Among all perpetrators, 546 (82.35%) self-reported at least one type of perpetration and of these, 71.43% (n = 390) did not contradict any of their partner’s reported victimization experiences, regardless of type. Conversely, 41.18% (n = 273) of identified perpetrators denied at least some of their perpetration—57.14% (n = 156) of these men self-reported at least one type of perpetration and denied at least one other; and 42.86% (n = 117) denied all perpetration.

Tabla 3

Crude and Adjusted Logistic Generalizing Estimating Actor-Partner Interdependence Models (APIM) Assessing the Odds of Denying Emotional IPV Perpetration Associated with Individual, Partner, and Relationship Characteristics in a Sample of Cisgender Male Couples (N = 527), United States, 2016-2017*

Note. Full model included all noted variables. Reference groups are noted in parens.

Abbreviations: odds ratio (OR); confidence interval (CI).

¹Centered at 18 years. ²Other race includes: native American or Alaskan native; native Hawaiian or other Pacific islander. ³Past 3 months; recreational drug use refers to any self-reported amphetamine, methamphetamine, hallucinogen, club drug, cannabis, analgesic narcotic, or other illicit substance use. ⁴Heterosexist Harassment, Rejection, and Discrimination Scale (HHRDS); possible values 18-84, higher values indicated higher experienced stigma. ⁵CESD-11 Iowa form; possible values 0-22, higher values indicate higher depressive symptoms. ⁶Partner value subtracted from actor value.

*p < .05.

Emotional IPV Perpetration

Bivariate associations between each type of IPV denial and individual-, partner-, and dyadic-characteristics are presented in Supplementary Table 1. Individual lower educational attainment and depression, and partner race/ethnicity (Hispanic, Black or African American), full- or part-time employment, higher experienced stigma, and depression were significantly, positively associated with denying emotional perpetration (p < .05). Emotional perpetration deniers tended to be older, and had lower experienced stigma and depression than their partners, whereas those who reported emotional perpetration were, on average, younger than their partners, with higher experienced stigma and depression scores (p < .0001).

Table 3 presents the logistic APIM results modeling the odds of emotional perpetration denial, accounting for dyadic interdependence. In the crude model, education was significantly associated with denial, but this effect became insignificant in the full model. In the crude and adjusted models, compared to those with Asian, multi-racial, or other race partners, those with non-Hispanic White, Hispanic, or Black/African American partners had significantly greater odds of denying emotional perpetration (p < .05). Comparatively, those with Black or African American partners had the highest adjusted odds of denial (OR = 4.90, 95% CI [1.17, 14.32]), closely followed by those with Hispanic partners (OR = 3.02, 95% CI [1.18, 7.75]). Those with unemployed partners had 61% lower adjusted odds of denying emotional perpetration (95% CI [0.18, 0.85]), compared to those with employed partners. In the crude models, dyadic differences in age and psychosocial scores exhibited stronger effects on perpetration denial than the corresponding partner variables, and were included in the full model. After adjusting for covariates, each 1-point increase in an individual’s depression score, relative to their partner’s score, resulted in 5% lower odds of denying emotional IPV perpetration (95% CI [0.90, 0.99]).

Tabla 4

Crude and Adjusted Logistic Generalizing Estimating Actor-Partner Interdependence Models (APIM) Assessing the Odds of Denying Monitoring/Controlling IPV Perpetration Associated with Individual, Partner, and Relationship Characteristics in a Sample of Cisgender Male Couples (N = 490), United States, 2016-2017

Note. Full model included all noted variables. Reference groups are noted in parentheses.

Abbreviations: odds ratio (OR); confidence interval (CI).

¹Centered at 18 years. ²Other race includes: native American or Alaskan native; native Hawaiian or other Pacific islander. ³Past 3 months; recreational drug use refers to any self-reported amphetamine, methamphetamine, hallucinogen, club drug, cannabis, analgesic narcotic, or other illicit substance use. ⁴Heterosexist Harassment, Rejection, and Discrimination Scale (HHRDS); possible values 18-84, higher values indicated higher experienced stigma. ⁵CESD-11 Iowa form; possible values 0-22, higher values indicate higher depressive symptoms. ⁶Partner value subtracted from actor value.

*p < .05.

Monitoring/controlling IPV Perpetration

According to bivariate analyses, monitoring/controlling perpetration deniers were significantly older, less likely to have reported recent binge drinking and/or recreational drug use, and had lower reported experienced stigma and depression than their counterparts who reported perpetrating monitoring/controlling IPV (p ≤ .01). In addition, monitoring/controlling perpetration deniers’ partners were significantly older, more depressed, and had higher experienced stigma than self-reported perpetrators’ partners (p < .001).

APIM results for denying monitoring/controlling IPV perpetration are presented in Table 4. In the crude models, age was positively significantly associated with the odds of denying monitoring/controlling perpetration. In contrast, individual depression scores and dyadic differences in experienced stigma and depression scores, past 3-month substance use, and a relationship length of 1 to 2 years (versus > 5 years) were significantly associated with reduced crude odds of monitoring/controlling perpetration denial. Dyadic differences in age, stigma, and depression were more strongly associated with monitoring/controlling perpetration denial than the partner-variables, and were included in the full model. After adjusting for covariates, the effects of relationship length, experienced stigma, and depression on denial were attenuated and no longer significant, and the effect of age was similar, but no longer significantly associated with denial. In the full model, substance users had 46% lower odds of denying monitoring/controlling perpetration than non-users (95% CI [0.32, 0.92]), and each 1-point increase in an individual’s depression score was associated with a 9% reduction in denial-odds (95% CI [0.83, 0.99]).

Physical/sexual IPV Perpetration

Physical/sexual IPV perpetration deniers had significantly lower reported experienced stigma and depression, according to bivariate analyses, and were less likely to have used recreational drugs/binge drinking in the previous 3-months, than their counterparts who reported perpetrating this type of IPV (p < .01). In addition, physical/sexual perpetration deniers’ partners were significantly older, but had significantly lower experienced stigma, than self-reported perpetrators’ partners (p ≤ .01). As with the other IPV types, physical/sexual perpetration deniers had lower depression scores than their partners, whereas those who reported perpetration tended to have higher depression scores than their partners (p < .0001).

Tabla 5

Crude and Adjusted Logistic Generalizing Estimating Actor-Partner Interdependence Models (APIM) Assessing the Odds of Denying Physical/Sexual IPV Perpetration Associated with Individual, Partner, and Relationship Characteristics in a Sample of Cisgender Male Couples (N = 490), United States, 2016-2017

Note. Full model included all noted variables. Reference groups are noted in parens.

Abbreviations: odds ratio (OR); confidence interval (CI); standard error (SE).

¹Centered at 18 years. ²Other race includes: native American or Alaskan native; native Hawaiian or other Pacific islander. ³Past 3 months; recreational drug use refers to any self-reported amphetamine, methamphetamine, hallucinogen, club drug, cannabis, analgesic narcotic, or other illicit substance use. ⁴Heterosexist Harassment, Rejection, and Discrimination Scale (HHRDS); possible values 18-84, higher values indicated higher experienced stigma. ⁵CESD-11 Iowa form; possible values 0-22, higher values indicate higher depressive symptoms. ⁶Partner value subtracted from actor value.

*p < .05.

Table 5 presents the logistic APIM results modeling the odds of denying perpetrating physical/sexual IPV, accounting for dyadic interdependence. In the crude models, substance use, lower partner education, and depression were significantly, negatively associated with denial, and having a non-Hispanic White partner (compared to a partner who self-identified as Asian, multi-racial, or other race) was positively, significantly associated with the odds of denial. In crude models, partners’ psychosocial characteristics were more strongly associated with denial than the corresponding dyadic difference variables; partner age, experienced stigma, and depression were included in the full model. After adjustment, substance use and partner education were no longer significantly associated with denial. However, each incremental increase in depression was associated with 10% lower odds of denying physical/sexual IPV perpetration (95% CI [0.83, 0.98]), after adjusting for covariates. Additionally, compared to those with Asian, multi-racial, or other race partners, the adjusted odds of denying physical/sexual perpetration among those with non-Hispanic White partners was 2.79 (95% CI [1.01, 7.68]); having a Hispanic or Black/African American partner was not significantly associated with physical/sexual perpetration denial in the adjusted model.

Discussion

Despite growing recognition of IPV and its impacts in SGM relationships, there are still substantial gaps in our understanding of how a partner sexual and gender identities shape IPV reporting (Marshall et al., 2011). In the current study, we identified notable discrepancies between partners’ victimization and perpetration reports, patterns of denial across IPV types, and individual and partner characteristics that were associated with denying perpetration. Specifically, large portions of individuals contradicted their partner’s reported victimization, particularly for physical/sexual violence. In this study sample, there were complex IPV reporting typologies, as much of the study’s perpetrators denied some forms of IPV while reporting perpetrating other type(s). This result suggests differential levels of reporting biases across IPV types. We also identified both individual- and partner-traits, emotional characteristics, and behaviors associated with IPV denial, indicating that perpetration denial and reporting is potentially dynamic, and relationship context-dependent.

Broadly, this study confirms the high prevalence of emotional, monitoring/controlling, and physical/sexual IPV in committed, long-term male partnerships in the U.S. Prevalence by IPV type followed previous findings in GBMSM populations, with emotional abuse the most prevalent and physical/sexual the least (Finneran & Stephenson, 2013b; M. Liu et al., 2021). Alone, these descriptive findings are an important reminder that there is an urgent need for GBMSM-targeted IPV prevention programs and interventions. Additionally, 365 couples were represented in the sample of identified perpetrators, 82% of which were complete dyads—i.e., both partners were categorized as perpetrators, indicating highly prevalent bidirectional IPV. Interestingly, the prevalence of bidirectional IPV by type mirrored overall prevalence trends. In mixed-gender couples, there is increasing recognition of abuse-congruence, but some debate as to whether female-perpetrated physical violence in bidirectional IPV-relationships should be labeled abuse or self-defense (Hine et al., 2022; Langhinrichsen-Rohling, 2010). This patriarchal, heteronormative framing is difficult to apply to male couples, where default assumptions about the aggressor’s gender may not apply, yet researchers are beginning to assert the need to further investigate the intricacies of gender, “mutual abuse,” and self-defense (C. Cannon, 2015; C. E. B. Cannon & Buttell, 2016; Langhinrichsen-Rohling et al., 2012; Messinger, 2018).

Overall, substantial proportions of identified perpetrators denied the abuse their partner reported experiencing—41% of perpetrators denied at least 1 type of IPV. Previous dyadic concordance studies in male-female couples have found a wide range of perpetration-underreporting or contradictory victimization/perpetration reports. On surveys, rates of discordant perpetration reports has been observed as low as 5.5 and as high as 83%, across a variety of samples (e.g., young unmarried adults; married veterans; polysubstance abusing couples) (Browning & Dutton, 1986; Cantos et al., 1994; Cui et al., 2005; Cunradi et al., 2009; Dutton & Hemphill, 1992; Heckert & Gondolf, 2000; Kuijpers, 2020; LaMotte et al., 2014; Langhinrichsen-Rohling & Vivian, 1994; Marshall et al., 2021; Marshall et al., 2011; Medina et al., 2004; Panuzio et al., 2006; Schafer et al., 2002; Wenger, 2015).

Denial patterns across IPV types exhibited substantial complexity in this study. In our study sample, identified perpetrators showed preference in terms of which types of IPV they reported and in which they denied engaging. Overall, perpetrators denied physical/sexual abuse to a greater extent than emotional or controlling/monitoring IPV. This confirms previous findings of differences in denial and underreporting by IPV type in mixed-gender couples (Caetano et al., 2009; Freeman et al., 2015; Kuijpers, 2020; Panuzio et al., 2006), and greater underreporting of more severe IPV, particularly sexual and physical abuse (Caetano et al., 2009; Freeman et al., 2015; Kuijpers, 2020; Panuzio et al., 2006). This likely reflects the greater stigma associated with physical/sexual violence compared to emotional abuse, and the social desirability and image management biases tied with reporting negative behaviors. This supports previous findings by Freeman et al. (2015), who noted that men who were more likely to intentionally manipulate people’s perceptions of them were also the most likely to underreport sexual aggression. Moreover, Heckert and Gondolf (2000) speculated that contextual factors can magnify image management’s role in denial; their research in a sample of court-mandated abusers found that men were more likely to deny their behavior when faced with potential social and/or legal ramifications, even if they had previously confirmed their aggressive behavior and/or their violence had been documented by police. However, the high levels of denial and underreporting in research settings and anonymized surveys, including the current study, suggest a knotty interplay between internal and external factors that motivate individuals to admit or deny their perpetration.

This study was the first of which we are aware to model IPV denial and reporting in male couples. These analyses identified some similarities, and further delineated differences, between physical/sexual, monitoring/controlling, and emotional IPV. Those with higher depression had significantly higher odds of reporting, as opposed to denying, perpetration, for both physical/sexual and monitoring/controlling IPV. Although individual depression was not a significant covariate in the emotional IPV model, higher depression scores relative to one’s partner was similarly associated with decreased denial. However, the mechanism and directionality underlying these relationships is unclear. It may be that acknowledging abusive behavior depresses mood and self-image and/or depression reduces one’s propensity for positive image management. This correlates with previous studies that have found that denial can be a distancing and self-defense mechanism used to protect one’s self-image (Barbaro & Raghavan, 2018; Scott & Straus, 2007). Alternatively, those without excessive depressive thoughts may have fewer negative perceptions of their relationship and behavior, and thus tend to underreport behavior. This hypothesis is supported by previous studies. Marshall et al. (2011) and LaMotte et al. (2014) both observed that high relationship satisfaction (both an individual and their partner’s) can lead to altered perceptions of events and IPV denial and underreporting (“satisfaction reporting bias”). In combination, these results suggest the relationship between mental health and perpetration denial may be modified or mediated by a relationship’s complete emotional context, as opposed to only individual mental health. Regardless, the significant associations between denial and modifiable characteristics, including depression and substance use indicate that perpetration denial and reporting may be dynamic. Previous research by Walsh and Stephenson (2022) also suggested that IPV reporting biases in cisgender male couples were not fixed. These results, in particular, present potential targets for intervention—perhaps addressing a relationship’s emotional context could improve the reliability of self-reported perpetration. Further analyses incorporating SGM’s relationship quality and satisfaction indices, as well as changes in denial over time, may shed light on this topic.

In addition to depression, substance use was negatively associated with monitoring/controlling perpetration denial. This aligns with previous Panuzio et al.’s (2006) finding of relatively high interpartner agreement on IPV occurrence from men enrolled in an alcohol treatment program and their female partners. In this study, the authors inferred that heavy-users might be motivated to report their behavior more accurately in help-seeking contexts. This may be relevant to the current study, as data was collected in the context of a couples-level intervention. Although participants were not aware of their study arm assignment when they took the baseline survey, it is possible that some participants viewed the study as a supportive context, which may have facilitated disclosure by perpetrators. Yet, previous research has also postulated that alcohol’s cognitive effects may lead to recall-related underreporting by both perpetrators and survivors (Armstrong et al., 2001; Medina et al., 2004), and others have suggested that involved parties may be less likely to hold perpetrators responsible for abuse occurring under the influence of alcohol, minimizing and excusing their behavior (LeJeune & Follette, 1994). More nuanced, further research into the relationships between substance use and IPV reporting and denial is warranted.

We also identified significant associations between emotional IPV perpetration denial and partner’s race and employment status. These associations appear to be independent of individual race and employment, as the unadjusted and adjusted effects were similar. Men with Hispanic, or Black/African American partners had significantly higher odds of denying their emotional perpetration, and men with non-Hispanic White partners had significantly higher odds of denying emotional and physical/sexual perpetration, compared to those with partners of another race (including multi-racial). Although research has noted that racial and ethnic minority SGM may experience higher IPV rates than White SGM (Stephenson et al., 2011), it is unclear how partner-race affects one’s own reporting. Minority stress related to sexuality-based discrimination has been associated with self-reported IPV perpetration (Stephenson & Finneran, 2017; Stephenson et al., 2011), and intersectional sexual/gender and racial/ethnic identities have been shown to compound IPV risk in some populations (Metheny & Stephenson, 2020). Although minority stress, measured as experienced homosexual stigma, was not significantly associated with IPV perpetration reporting in the current study, the identified associations between denial and partner race/employment in this sample of GBMSM may reflect some aspect(s) of internalized structural stress and bias. Additional research into the relationships between IPV reporting, discrimination, and structural health determinants is necessary.

This study was limited in a few ways. As an adscititious analysis to the parent CHCT intervention, we were unable to explore potentially important independent predictors of perpetration denial, such as image management, for which data was not collected. In addition, research has noted that IPV perpetration denial and data reliability are not necessarily static within individuals or couples, making it difficult to generalize results from cross-sectional studies (Abramsky et al., 2022; Loxton et al., 2019; Pachana et al., 2011; Walsh & Stephenson, 2022). This cross-sectional study also did not collect data on IPV antecedents or context, and therefore we were unable to differentiate between defensive and offensive violent acts in this study, or describe more complex IPV typologies. As this study population was an opportunistic sample from a parent study with strict inclusion criteria, these results also may not be broadly generalizable, particularly to other types of relationships or contexts with distinct IPV risks and reporting tendencies (e.g., short-term relationships, different sexual and gender identity-couples), or for severe IPV, as the study sample excluded those who reported feeling unsafe in their relationship. This may be a particularly important point, as research has noted links between IPV severity and victimization and perpetration minimization/denial (Helfritz et al., 2006; Kropp & Gibas, 2020). In addition, as with all unobserved behaviors, it is impossible to know the quaesitum of IPV prevalence. It is likely that there was measurement error in both perpetration and victimization related to the survey’s IPV instrument and its individual items (Stephenson et al., 2019). Donovan and Barnes (2020) suggested that not all “abusive” behaviors were universally problematic (particularly non-physical actions), noting that IPV survey items can be open to interpretation related to behavioral intent and relationship context. Yet, it is difficult to speculate as to the extent or directionality of reporting biases, however, given the possible combinations of differential and non-differential victimization and perpetration under- and over-reporting within and across IPV types and individual behaviors. In addition to denial, there is significant evidence that many perpetrators minimize the severity and/or frequency of their abuse (Morrison et al., 2021; Scott & Straus, 2007). In the current study, we investigated denial of each IPV type as a dichotomous outcome, which may have masked important distinctions between denial, minimizing, and reporting. Researchers have noted the importance of granularity in IPV data reliability and interpartner agreement (Kuijpers, 2020; Szinovacz, 1983; Walsh & Stephenson, 2022); more granular research may reveal nuances in perpetration reporting tendencies. Finally, there may be unaccounted for-confounding present in the current results. As always, however, these limitations present avenues upon which to build further research.

This study’s exploration into patterns and correlates of IPV perpetration denial in cisgender, primarily homosexual, male couples is a small step towards a clearer picture of IPV and abuse in GBMSM, a concerningly under-researched public health issue. The results of this study illustrate the importance and utility of dyadic data in developing a greater understanding of IPV’s complexities, and should spur continuing efforts to collect complete within-relationship IPV data. Moreover, this study highlights differences in reporting and denial across IPV types, supporting the need to develop type- and role-specific IPV measurement tools for use in SGM populations. Broadly, the current results suggest that continued research into male couples’ IPV is needed, particularly research that collects data from both partners on their experienced and enacted abuse, as well as information related to context and timing, relationship quality, emotional well-being, demographics, and cultural and structural contexts. Improving our global understanding, and the visibility, of IPV in SGM relationships will lead to increased measurement validity and reliability, more accurate prevalence and incidence estimates, and ultimately more effective interventions.