Internalized Homophobia and Relationship Quality among Same-Sex Couples: The Mediating Role of Intimate Partner Violence

Abstract

Based on a geographically and socioeconomically diverse sample of 144 same-sex couples and using a dyadic approach (i.e., the Actor-Partner Interdependence Mediation Model [APIMeM] with interchangeable dyads), this study examined the association between internalized homophobia and same-sex relationship quality, and also tested the potential mediating role of intimate partner violence perpetration in this association. Results indicated that individuals’ own and their partners’ psychological violence perpetration mediated the negative associations from individuals’ own internalized homophobia to individuals’ own and their partner’s relationship quality. Such findings contribute to the understanding of mechanisms underlying the harmful effects of sexual minority stressors for same-sex relationship well-being. Implications for interventions were also discussed.

The increasing visibility of same-sex couples has challenged researchers to obtain more scientific understanding of the factors contributing to same-sex relationship well-being (Peplau & Fingerhut, 2007; Umberson, Thomeer, Kroeger, Lodge, & Xu, 2015). Given their historically disenfranchised status in various social spheres, same-sex couples often experience increased vulnerabilities for relationship well-being (Frost, 2011). Notably, decades of research on sexual minority couples has particularly focused on the implications of various stressors for same-sex relationship well-being (Doyle & Molix, 2015; LeBlanc, Frost, & Wight, 2015; Rostosky & Riggle, 2017).

Among various stressors, internalized homophobia (IHP), which is defined as “the gay person’s direction of negative social attitudes toward the self, leading to a devaluation of the self and resultant internal conflicts and poor self-regard” (Meyer & Dean, 1998, p. 161), has long been considered as one of the most salient and proximal predictors of same-sex relationship well-being (Cao, Zhou, Fine, Liang, Li, & Mills-Koonce, 2017). An expanding body of research has consistently identified significant associations between IHP and various dimensions of same-sex relationship well-being, including satisfaction, intimacy, stability, commitment, aggression, and conflicts (e.g., Feinstein, McConnell, Dyar, Mustanski, & Newcomb, 2018; Frost & Meyer, 2009; Meyer & Dean, 1998; Mohr & Daly, 2008; Otis, Rostosky, Riggle, & Hamrin, 2006).

Yet, some limitations in the existing research are worth noting. First, the available research has almost exclusively focused on the direct association between IHP and same-sex relationship well-being. To make further sense of the implications of IHP for same-sex relationship well-being, researchers may need to take a process perspective to better delineate mechanisms underlying these associations. Theoretically, IHP may contribute to individuals’ diminished self-esteem, attachment insecurity, fears of intimacy, and doubts toward oneself and others (Meyer & Dean, 1998). Individuals with higher levels of IHP are more likely to internalize the prevailing societal views of the dysfunctions of same-sex relationships (i.e., highly instable, low committed) as their personal expectations. Thus, these individuals may avoid establishing long-term and highly-invested relationships in order to protect themselves from potential losses and threats (Otis et al., 2006). As indicated in prior studies based on different-sex samples, some of aforementioned factors (e.g., diminished self-esteem, attachment insecurity, low commitment) may predispose individuals to the perpetration of violence when interacting with intimate partners (for review, see Capaldi, Knoble, Shortt, & Kim, 2012). Thus, it is warranted to expect that IHP will increase intimate partner violence and, in turn, diminish relationship well-being among same-sex couples.

Second, one of the most defining characteristics of couple relationship is the interdependence between partners, which highlights the importance of utilizing dyadic approaches in couple research. Nevertheless, the vast majority of the existing studies on the links between IHP and same-sex relationship outcomes have collected data from only one partner in a couple and/or analyzed data from an individual rather than a dyadic perspective (for exceptions, see Feinstein et al., 2018; Otis et al., 2006; Totenhagen, Ryndall, & Lloyd, 2018). Moreover, even when both partners’ data are available, researchers still face a challenge when examining the association between IHP and same-sex relationship well-being from a dyadic perspective. That is, it is difficult to distinguish one partner from the other simply based on their sex in a same-sex relationship (i.e., the interchangeable nature of same-sex dyads; Olsen & Kenny, 2006; Sadler, Ethier, & Woody, 2011). Efforts appropriately addressing this issue with more rigorous approaches are pressing (i.e., the Actor-Partner Interdependence Model with interchangeable dyads).

Lastly, the samples used in prior research in this field were often restricted in characteristics such as geographical locations (i.e., from a single or just a few U.S. states), union types (i.e., primarily or exclusively cohabitating couples without any legal documents), race/ethnicity (i.e., predominantly or exclusively Non-Hispanic White), and socioeconomic status (e.g., primarily or exclusively middle-class). As some researchers stated (e.g., Karney & Bradbury, 2005), such samples are problematic when examining stress issues because: (a) a rather narrow range of stressors may be represented among these couples; and (b) couples in such samples are likely to possess more resources that can help them cope with stressors and thus may attenuate the associations between stress and relationship outcomes. Thus, it seems critical to revisit the association between IHP and same-sex relationship well-being in more geographically and socioeconomically diverse samples.

To address the aforementioned limitations, in the present study we seek to examine the association between IHP and same-sex relationship quality and also test the potential mediating role of intimate partner violence in this association with a dyadic approach and using data from a more geographically and socioeconomically diverse sample of same-sex couples. Notably, given the indistinguishability of two partners in a same-sex couple, we use a more rigorous statistical strategy in analyses (i.e., the Actor-Partner Interdependence Mediation Model [APIMeM] with interchangeable dyads; Ledermann, Macho, & Kenny, 2011; Olsen & Kenny, 2006; Sadler et al., 2011).

Literature Review

Theoretical Framework

The currently proposed model (as depicted in Figure 1) is informed by the vulnerability-stress-adaptation (VSA) model (Karney & Bradbury, 1995). Although the VSA model was originally developed based on studies of different-sex couples, an emerging (yet still limited) body of research has indicated the potentials to utilize the VSA model in the examinations of same-sex relationships (Feinstein et al., 2018; Kurdek, 2006; Totenhagen et al., 2018). One of the core propositions in the VSA model is particularly relevant for the present study. That is, vulnerabilities, which refer to enduring traits that each partner brings into the couple relationship, shape subsequent couple relationship well-being via affecting the adaptive processes between partners (Karney and Bradbury, 1995).

Figure 1.

IHP, IPV, and relationship quality among same-sex couples.

Note. IHP = Internalized Homophobia, IPV = Intimate Partner Violence. Presented are standardized coefficients. To note, (a) physical IPV perpetration included in the pathway analyses is the dichotomized scale score; (b) IHP, psychological IPV perpetration, and relationship quality were transformed scores; (c) pathways/correlation lines with p > .05 are depicted in dash, gray lines; and (d) pathways/correlation lines with p < .05 are depicted in solid, black lines. * p < .05, ** p < .01, *** p < 001 (two-tailed).

In the present study, IHP (i.e., same-sex person’s negative attitudes toward the self, devaluation of the self, and the associated inner conflicts and poor self-regard; Meyer, 2003; Meyer & Dean, 1998) can be considered as an individual vulnerability. Intimate partner violence (IPV), which refers to the aggressive behaviors between partners in couple interactions (Heyman, Foran, & Wilkinson, 2010), is an indicator for couple adaptive processes. According to the aforementioned proposition from the VSA model, it is expected that IHP (as a personal vulnerability) is negatively associated with same-sex relationship quality (as an indicator of couple relationship well-being) through increased levels of IPV (as an indicator of couple adaptive processes).

Empirical Studies

Prevalence of IPV in same-sex community

IPV includes different forms of psychologically (e.g., calling the partner’s name and demeaning the partner) and physically (e.g., pushing, grabbing, and slapping the partner) aggressive behaviors between partners in a close relationship (Heyman et al., 2010). IPV is not uncommon in the U.S. Based on data from nationally representative samples of general population (i.e., the National Intimate Partner and Sexual Violence Survey), the lifetime prevalence of psychological IPV victimization was 36.4% among women and 34.3% among men, and the lifetime prevalence of physical IPV victimization was 30.6% among women and 31.0% among men (Smith, Zhang, Basile, Merrick, Wang, Kresnow, & Chen, 2018).

For the prevalence of IPV in same-sex population, data from a nationally representative sample of U.S. adults indicated that individuals in same-sex relationships victimized higher levels of IPV during the lifetime than did individuals in different-sex relationships (Messinger, 2011). Specifically, the lifetime prevalence of psychological IPV victimization was 65.5% among males in same-sex relationships and 44.0% among females in same-sex relationships; the lifetime prevalence of physical IPV victimization was 33.3% among males in same-sex relationships and 25.0% among females in same-sex relationships (Messinger, 2011). Moreover, by summarizing 14 studies on U.S. community samples of female in same-sex relationships, a meta-analytic review found that the lifetime prevalence of IPV victimization was 43% for psychological IPV and 18% for physical IPV (Badenes-Ribera, Frias-Navarro, Bonilla-Campos, Pons-Salvador, & Monterde-i-Bort, 2015). In addition, and varying as a function of the recall period from the past 6 months to the lifetime, prior studies found that the prevalence of IPV victimization among male in same-sex relationships ranged from 5.4% to 73.2% for psychological IPV and from 11.8% to 45.1% for physical IPV (Finneran & Sephenson, 2013a).

IHP and IPV

As stated earlier, it is theoretically warranted that higher levels of IHP are linked to a series of risk factors for IPV. Empirically, a handful of studies using cross-sectional data from same-sex partners (rather than couples) have provided preliminary evidence supporting that individuals’ own IHP may be an important precursor for their own IPV perpetration (i.e., the actor association) and their partner’s IPV perpetration (i.e., the partner association). Based on 272 females’ reports on their experiences in the same-sex couple relationships, Balsam and Szymanski (2005) found that higher levels of participants’ IHP was associated with higher levels of their own and their partners’ psychological IPV perpetration. Such findings were replicated in a sample of 220 females in same-sex relationships (Lewis, Milletich, Derlega, & Padilla, 2014). Likewise, based on data from 2,368 males in same-sex relationships, Finneran et al. (2012) also found that participants’ higher levels of IHP were related to higher levels of their own and their partners’ physical IPV perpetration.

IPV and same-sex relationship quality

Previous studies on different-sex relationships have conducted dyadic analyses and found robust negative associations from one partner’s IPV perpetration to their own and their partner’s relationship quality (Caldwell, Swan, & Woodbrown, 2012; Hammett, Lavner, Karney, & Bradbury, 2017; Shortt, Capaldi, Kim, & Laurent, 2010). In contrast, research with samples of same-sex couples remains sparse, and the available few studies have been predominantly based on data from one partner in a couple (Kelley, Lewis, & Mason, 2015; Stephenson, Rentsch, Salazar, & Sullivan, 2015). Using data from 819 females in same-sex relationships, Kelley and colleagues (2015) identified a middle-sized, negative association between participants’ partner’s psychological IPV perpetration and participants’ own relationship quality (i.e., the partner association). Using data from 525 males in same-sex relationships, Stephenson et al. (2015) found that higher levels of partners’ psychological or physical IPV perpetration were related to lower levels of participants’ own relationship quality (i.e., the partner association).

The Present Study

To extend the existing literature, we conducted an APIMeM to examine (a) the associations between IHP and same-sex relationship quality, and (b) the mediating roles of IPV in these associations. Notably, a series of potential covariates were controlled for due to their associations with the key study variables: general life stress, relationship length, couple type based on sex, race/ethnicity, income status, union status, parental status, and state-level sociocultural climate. In particular, general life stress (i.e., stressful life events) have been identified as risk factors of IPV in the general population (Capaldi et al, 2012). As same-sex couples experienced both general life stress (i.e., stressful life events) and LGBT-specific stressors, we controlled for general life stress to identify the unique roles played by IHP on same-sex relationship quality. Further, longer relationship length has been found to be associated with the lower levels of IPV perpetration (Rickert, Wiemann, Harrykissoon, Berenson, & Kolb, 2002). IPV perpetration may vary across genders (Capaldi et al, 2012). Racial minority status and low socioeconomic status have been demonstrated to be related to higher levels of IPV in same-sex community (Edwards, Sylaska, & Neal, 2015). Being in marriage or in registered domestic partnership have been shown to be associated with lower levels of IHP (Riggle, Rostosky, & Horne, 2010). Having child (Riggle et al., 2010) and living in more liberal states (Goldberg & Smith, 2011) have been found to be associated with higher levels of well-being among individuals in same-sex relationships.

While controlling for aforementioned covariates, we tested the following hypotheses:

• H1: Individuals’ own IHP will be negatively related to their own and their partner’s relationship quality.

• H2: Individuals’ own IHP will be positively related to their own and their partner’s psychological and physical IPV perpetration.

• H3: Individuals’ own psychological and physical IPV perpetration will be negatively related to individuals’ own and their partners’ relationship quality.

• H4: Individuals’ own and their partner’s psychological and physical IPV perpetration will mediate the negative associations between individuals’ own IHP and either individuals’ own or their partner’s relationship quality.

Method

Participants and Procedures

Data in the current sample were derived from a larger project examining the formation and functions of families headed by sexual minority couples. The data in the larger project have not been published, yet the principal investigators can provide data upon request. In 2014 and 2015, couples were recruited through online advertisements on Craigslist and Facebook as well as snowballing that were targeted at all 50 states in the U.S. To be eligible, couples should be (a) in a cohabiting romantic relationship with a same-sex partner or in a cohabiting relationship in which at least one partner’s identification was transgender or gender non-conforming; and (b) above the age of 18 years. Although researchers encouraged both partners in the relationship to participate in the study, this was not a mandatary requirement, because the larger project aimed to examine not only couple and family functioning among sexual minority population but also individual well-being for those who were in LGBTQ relationships. If only one partner within a relationship agreed to participate, we included him/her in the study and obtained consent from himself/herself only. When both partners were willing to take the survey, we included two partners in the study and obtained consent from each of them. Ultimately, 212 couples living in 48 states across the U.S. and also the District of Columbia participated in the project.

Procedures of this study were approved by its home university institutional review board. Data were collected via Qualtrics, as it was flexible and economic to collect data from widely dispersed areas. For every participant included in the larger project, research assistants sent a secure email that contained a brief instruction, a unique ID number, and an individual link to the online survey. The ID number was 9-digit study ID that was created by research assistants in advance. In couples where two partners agreed to participate, the first 8 digits were identical so that researchers can pair them, and the last digit was randomly assigned as 0 for one partner and 1 for the other. The individual link directed participants to a secure website on which they were required to independently (a) enter in the 9-digit ID number, (b) sign the online consent form, and (c) complete a series of online questionnaires. Every participant received a $10 Amazon e-gift as compensation in about one week after the completion of the survey.

As the primary purpose of the present study was to examine same-sex couple relationship well-being, we excluded data from couples (a) involving transgender and gender non-conforming partners, or (b) in which only one partner responded to the survey. Ultimately, the present study included 288 partners in 144 same-sex couples. We conducted a multivariate analysis on key study constructs and demographic variables to examine the potential differences between 288 partners in the current the study and those in couples in which only one partner responded to the survey. One small-sized difference emerged. In comparison to those in couples in which only one responded to the survey, partners in the present study reported less frequent physical IPV perpetration [M _included = 1.31 versus M _{not included} = .54; F = 3.87, adjusted p = .050 (two-tailed level); partial η² = .015].

For the 288 partners, the mean of ages was 34.24 years old (SD = 9.90), and the mode of education was “some college but without degree.” For 144 couples, the mean of relationship length was 5.24 years (SD = 5.17). For sex, 75.7% were same-sex female dyads. For income status, 32.0% of couples were in low-income status (i.e., income-to-needs ratio ≤ 2, based on data from Census Bureau of U.S. https://www.census.gov/hhes/www/poverty/data/threshld/). For parental status, 62.5% of couples had no child. For union status, 45.8% of couples were legally married or at least in registered domestic partnership or civil unions. For race/ethnicity, 29.6% of couples involved at least one partner of racial/ethnic minority. For state-level sociocultural climate, 41.0% of couples were living in socio-culturally liberal states [i.e., states that fell in the lowest third of percent voting republican (see http://library.cqpress.com/elections/) and were early adopters (i.e., before Oct. 2014) of same-sex marriage laws (see http://www.ncsl.org)]. We also displayed the demographic information respectively for male and female dyads in Table 1.

Table 1.

Demographic information for male and female dyads (N = 144 couples)

Demographic variables	Female dyads (n = 109 couples)	Male dyads (n = 35 couples)	t-tests
Mean
Age (in years)	33.10	37.05	t = −2.17*
Relationship length (in years)	4.95	6.16	t = −.99
Valid Percentage
Living in more liberal states	43.1%	34.3%	t = .94
Having child(ren)	43.1%	20.0%	t = 2.77**
Involving at least one racial/ethnic minority partner	32.4%	20.6%	t = 1.41
Low-income status	40.8%	7.4%	t = 3.56***
Legally married, registered domestic partnership or civil unions	46.0%	51.4%	t = −3.33***

Note.

^*p < .05

^**p < .01

^***p < .001 (two-tailed).

Measures

IHP

A modified version of the Lesbian Internalized Homophobia Scale (LIHS; Szymanski & Chung, 2001) was used to assess IHP perceived by same-sex partners. Despite that the LIHS was developed originally among females in same-sex relationships, prior studies have indicated relatively high levels of reliability (i.e., Cronbach’s α = .90) and criterion validity (i.e., middle-sized positive correlations with substance abuse) of LIHS across LGBTQ populations (Amadio, 2006; Amadio & Chung, 2004). For the purpose of the larger project, all the items were revised to assess internalized homophobia for all lesbian/gay/trans* individuals (i.e., using “lesbian/gay/trans” rather than “lesbian” in each item statement). An example item of the modified scale was “ I dislike myself for being attracted to other women/men/trans*.” On each item, participants were asked to indicate their agreement with statements on a 7-point Likert scale ranging from 1 (strongly agree) to 7 (strongly disagree). After reversing scores for thirteen negative/inverse items, we followed procedures in existing studies and calculated the mean scores of all the items (Morandini et al., 2015; Szymansk & Kashubeck-West, 2008). Higher scores indicated higher levels of IHP. Cronbach’s α in the present study was .91.

Notably, the original version of LIHS included 52 items and 5 subscales: connection with the lesbian community (CLC; 13 items), publication identification as a lesbian (PIL; 16 items), personal feeling about being a lesbian (PFL; 8 items), moral and religious attitudes toward lesbians (MARTL, 7 items), and attitudes toward other lesbians (ATOL; 8 items). In the larger project, the 24 items from the PIL and PFL subscales were used given the following considerations. First, as the original scale developers reported, neither MARTL and ATOL were significantly correlated with self-esteem, suggesting the low criterion validity of these two subscales (Szymanski & Chung, 2001). Thus, also in line with existing studies (e.g., Morandini, Blaszczynski, Dar-Nimrod, & Ross, 2015), the research team trimmed these two subscales off to shorten the length of entire survey used in the larger project. Second, and also as the original scale developers presented, the CLC subscale due to its less robustness in the factor analyses (i.e., an indicator for low construct validity; for details, see Szymanski & Chung, 2001; Szymansk & Kashubeck-West, 2008). As such and aiming to shorten the length of entire survey used in the larger project, the CLC subscale was not used either.

IPV

The Conflict Tactics Scale-Couple Form Revised (CTS-CF-R; Straus, Hamby, Boney-McCoy, & Sugarman, 1996) was used to assess psychological IPV perpetration and physical IPV perpetration in same-sex relationships. Existing studies based on same-sex sample suggested that CTS-CF-R has relatively high levels of internal reliability (i.e., Cronbach’s αs > .80; Kelley et al., 2015) and construct validity (i.e., fit indices for confirmatory factor analyses were χ² = 33.80 (p = .01), CFI = .95, and RMSEA = .08; Matte & Lafontaine, 2011). Moreover, prior studies using CTS-CF-R identified middle-sized, negative associations between IPV and same-sex couple relationship well-being, which indicates the criterion validity of CTS-CF-R in same-sex populations (Kelley et al., 2015).

The psychological IPV subscale contained 6 items (e.g., Shout at him/her/you), and intimate physical IPV subscale included 9 items (e.g., Beat him/her/you up). For each item, participants were asked to report whether specific violent behavior never occurred, occurred once, twice, 3–5, 6 –10, 11–20, or more than 20 times in past 12 months (i.e., a 7-category scale). To calculate the frequency of psychological and physical IPV in past 12 months, Straus (1995) suggested to convert the above 7 categories into single digit [i.e., never =0 as 0, once = 1, twice = 2, 3–5 times =4, 6–10 times = 8 (i.e., median of the range), 11–20 times = 15 (i.e., median of the range), and more than 20 times = 25 (i.e., median of the range)]. We then summed up the single digit of items within each scale to reflect how many times in total the psychological IPV or physical IPV occurred during past 12 months (Straus, 1995). Higher scores indicated more frequent violent behaviors. Cronbach’s αs in the present study was .79 for psychological IPV and .72 for physical IPV.

To note, on each item of CTS, participants need to report how many times themselves and their partners perpetrated specific violent behaviors. In both prior studies (for a meta-analysis, see Simpson & Christensen, 2005) and the current study (ICC = .61 for psychological IPV and .35 for physical IPV), low-to-moderate levels of consistency emerged between focal participants’ self-report perpetration and their partners’ report on focal participants’ perpetration. Moreover, given the self-serve bias and the social desirability bias, focal participants’ self-report perpetration is typically underreported whereas their partners’ report on focal participants’ perpetration tends to be overreported (for reviews, see Chan, 2011; Simpson & Christensen, 2005). To correct these biases and more accurately capture the frequency of IPV perpetration in same-sex couple relationships, we followed the procedures in existing studies (e.g., Bradley, Drummey, Gottman, & Gottman, 2014) and used the average scores of participants’ self-report perpetration and their partners’ report on focal participants’ perpetration to represent the focal participants’ IPV perpetration.

Relationship quality

The 6-item, unidimensional Quality Marriage Index Scale (QMI; Norton, 1983) was used to assess relationship quality. As a global measurement of marital quality, the QMI is robust against the interpretation problems that may arise in the omnibus measures of marital quality (e.g., Dyadic Adjustment Scale; Bradbury, Fincham, & Beach, 2000). Moreover, prior studies have demonstrated high levels of reliability of QMI in same-sex population (i.e., Cronbach’s αs > .90; see Graham & Barnow, 2013).

The first 5 items asked partners to indicate their agreement with statements such as “My relationship with my partner makes me happy.” on a 7-point Likert scale ranging from 1 (very strong disagreement) to 7 (very strong agreement). The last item asked partners to indicate how happy they were in their relationship when all things were considered on a 10-point Likert scale from 1 (very unhappy) to 10 (perfectly happy). For the purpose of the present study, all items were revised to assess relationship quality for same-sex couples (i.e., using “relationship” or “partner” rather than “marriage” or “spouse” in each item statement). Mean scores were calculated and used in analyses, with higher scores indicating higher levels of relationship quality. Cronbach’s α in the present study was .94.

Covariates

A 10-item, abbreviated stress life event experience scale was used to assess general life stress encountered by each partner. It is a modified version of the Life Experiences Survey (LES; Sarason, Johnson, & Siegel, 1978). Each item described a specific life event (e.g., serious illness or injury of a close friend or family member). For each item, participants were first asked to indicate whether the event occurred or not during the last 6 months. If the event occurred, participants were asked to further indicate whether it was a good or bad experience and the extent to which it affected their life on a 4-point Likert scale from 0 (no effect) to 3 (great effect). To evaluate how bad the life experiences were for each participant, the mean score of weighted negative events was calculated and used in analyses. That is, we first summed up the number of events rated as “bad” and then summed up the weights (i.e., the effect of the event on life) associated with each item rated as “bad”, and then the sum of weights of “bad” events was divided by the total number of the “bad” events.

A series of binary variables were created for couple type based on sex, race/ethnicity, income status, parental status, union status, and the state-level sociocultural climate. For couple type based on sex, same-sex female dyads were coded as 0, and same-sex male dyads as 1. For couple type based on race/ethnicity, 0 indicated that both partners in a couple were Non-Hispanic White and 1 indicated at least one partner in a couple was racial/ethnic minority. For couple type based on income status, 0 indicated low-income status and 1 indicated non-low-income status. For couple type based on parental status, couples having no child were coded as 0, and couples having at least one child as 1. For couple type based on union status, couples cohabitating without any legal recognition were coded as 0, and couples legally married or in registered domestic partnership or civil unions as 1. For couple type based on state-level social cultural climate, 0 indicated socio-culturally conservative states, and 1 indicated socio-culturally liberal states. To note, state-level social cultural climate was a self-developed item, and research team of the larger project selected the two indexes on the basis of prior studies (i.e., the index for how supportive the state law is for same-sex marriage and families; Goldberg & Smith, 2011; Rostosky, Riggle, Horne, & Miller, 2009) and group discussion (i.e., the index for the proportion of individuals voting for republican). Validity of the self-developed item on state-level social cultural climate was demonstrated by our finding that couples in liberal states experienced lower levels of IHP than couples in conservative states (see Table 3).

Table 3.

Correlation analyses using transformed scores of key study constructs (N = 144 couples)

	IHP	Psychological IPV perpetration	Physical IPV perpetration a	Relationship quality
Key study constructs
IHP (WPA)
IHP (BPA)	.21**
Psychological IPV perpetration (WPA)	.15*
Psychological IPV perpetration (BPA)	.17**	.83***
Physical IPV perpetration (WPA)	.11†	.47***
Physical IPV perpetration (BPA)	.09	.44***	.66***
Relationship Quality (WPA)	−.25***	−.46***	−.28***
Relationship Quality (BPA)	−.18**	−.46***	−.29***	.65***
Covariates
General life stress (WPA)	.06	.11*	.09	−.02
General life stress (BPA)	.03	.09	.06	−.05
Relational length	.04	−.01	−.08	−.04
Couple type based on sex ref = female, b	.44	−2.73**	−.84	1.21
Couple type based on race/ethnicity ref = white, b	1.80†	1.11	.92	−1.52
Income status ref = non-low-income status, b	−.15	.80	.45	.20
Union status ref = legally married or registered	.69	−.55	1.13	.27
Parental status ref = having no child, b	−1.14	.41	1.68	−3.55**
State-level sociocultural climate ref = liberal, b	1.89†	.89	.62	−.25

Note. IHP = Internalized Homophobia, IPV = Intimate Partner Violence, and ref = Reference group. Bold values represented intraclass ICCs. The value .15 was the association between two partners’ IHP within a same-sex dyad. Non-bolded values represented bivariate ICCs. WPA refers to within-partner associations, and BPA refers to between-partner associations. For example, .15 in the first column represented the association between individuals’ IHP and their own psychological IPV perpetration, and .17 in the first column represented the association between individuals’ IHP and their partners’ psychological IPV perpetration. Significance was calculated using z scores with adjusted standard errors (Kenny et al., 2006).

^†p < .10

^*p < .05

^**p < .01

^***p < .001 (two-tailed).

^aPhysical IPV perpetration in the table was dichotomized.

^bFor binary covariates covariance, we conducted t-test on key study constructs and displayed the independent-t in the table.

Analytic Strategies

Path analyses were conducted in Mplus 7.4 and the actor-partner interdependence mediation model (APIMeM; Ledermann et al., 2011) was used. Missing values were handled with full information maximum likelihood estimation method (FIML) (Acock, 2005). Two partners in a same-sex couple should be regarded as “interchangeable” from each other, as their sex does not vary within a couple (Kenny, Kashy, & Cook, 2006; Olsen & Kenny, 2006). Theoretically, within interchangeable dyads, two partners in a same-sex couple should have the same “population mean and variance on the predictor variable, the same actor effect, the same partner effect, the same intercept on the outcome variable, and the same error variance” (Sadler et al., 2011, p. 121). As such, we added model constraints to fix these parameters of two partners to be equal (e.g., Olsen & Kenny, 2006; Sadler et al., 2011).

To test specific indirect associations via each mediator, multiple mediators (i.e., psychological IPV perpetration and physical IPV perpetration of both partners) were simultaneously included in a single model. Indirect effects were estimated using bootstrapping, as this nonparametric method does not assume normal distribution of indirect effects and can therefore adjust inflated type I and type II errors (Preacher & Hayes, 2008). The bias-corrected bootstrapped Confidence Intervals (CIs) were based on 1,000 resamples. Conclusions regarding the statistical significance of indirect pathways were on the basis of 95% bias-corrected bootstrapped CIs around the unstandardized indirect associations.

Given that relationship length, income status, and union status were not significantly associated with the key study constructs in the represent study (seen in Table 3), we trimmed them off from the final analyses to increase the model parsimoniousness. As displayed in Figure 1, six covariates were retained to be controlled for (i.e., general life stress for both partners as well as couple type based on sex, race/ethnicity, parental status, and state-level sociocultural climate). Equality constraints were also added on (a) the paths from the covariates to each partner’s relationship quality, and (b) covariance between covariates and either IHP or IPV perpetration.

Results

Preliminary Analyses

Prevalence of IPV perpetration

For the prevalence of psychological IPV perpetration in the present study, 94.1% of the 288 partners perpetrated psychological IPV during the past 12 months. Moreover, 89.6% of the 144 couples experienced bidirectional psychological IPV perpetration (i.e., two partners within the relationship both perpetrated psychological IPV), and 3.5% of the 144 couples experienced unidirectional psychological IPV perpetration (i.e., one partner within the relationship perpetrated psychological IPV yet the other not).

For the prevalence of physical IPV perpetration in the present study, 23.6% of the 288 partners perpetrated physical IPV during the past 12 months. Moreover, 17.4% of the 144 couples experienced bidirectional physical IPV perpetration (i.e., two partners within the relationship both perpetrated physical IPV), and 12.5% of the 144 couples experienced unidirectional physical IPV perpetration (i.e., one partner within the relationship perpetrated physical IPV yet the other not). The 12-month prevalence of psychological and physical IPV perpetration in the present study is higher than those in previous studies based on community sample of general population (Elliott, Cunningham, Colangelo, & Gelles, 2011; Straus, 2017) and same-sex population (Edwards & Sylaska, 2013; Finneran & Stephenson, 2013b).

Distribution and transformation of key study variables

As displayed in Table 2, the original scores of IHP, psychological IPV perpetration, and physical IPV perpetration were positively skewed (i.e., skewness > 1); the original scores of relationship quality were negatively skewed (i.e., skewness < −1; Jain, 2018). To reduce the errors along with the skewed distribution, log transformation was conducted on IHP, psychological IPV perpetration, and physical IPV perpetration; and reciprocal transformation was conducted on relationship quality (for detailed guideline, see Osborne, 2005).

Table 2.

Descriptive analyses and normality diagnosis (N = 144 Couples)

	IHP	Psychological IPV perpetration	Physical IPV perpetration	Relationship quality
Original Scores
Full range on scale	1–7	0–150	0–225	1–7.5
M [Observed range]	2.02 [1–5.63]	17.39 [0, 91.50]	.45 [0, 15.50]	5.75 [1–6.67]
SD	.86	19.25	1.46	1.26
Skewness	1.18	1.65	6.25	−1.95
Kurtosis	1.39	2.59	51.06	3.49
Transformed Scores
Skewness	.33	−.26	2.73	−.31
Kurtosis	−.65	−.71	7.91	−1.28

Note. IHP = Internalized Homophobia, IPV = Intimate Partner Violence.

After transformations, the distributions of IHP, psychological IPV perpetration, and relationship quality became fairly normal (i.e., Table 2, skewness and kurtosis fall into the range from −.50 to .50; Jain, 2018). As such, we used the transformed scores of IHP, psychological IPV perpetration, and relationship quality in subsequent analyses. However, the transformed scores of physical IPV perpetration were still highly skewed (i.e., see Table 2, skewness > 1; Jain, 2018). To this end, we followed procedures in prior studies (e.g., Elliott et al., 2011) and recoded the original scale score of physical IPV perpetration into a dummy code (0 = no physical IPV was reported during 12 months, 1 = physical IPV was reported during past 12 months). The dichotomous score of physical IPV perpetration was used in analyses.

Correlation analyses

To note, as two partners in interchangeable dyads were arbitrarily assigned, bivariate correlations should be calculated via pairwise approach (Griffin & Gonzalez, 1995; Kenny et al., 2006). The associations between both partners’ reports of the same variable can be termed as intra-class correlation coefficients (ICCs), and the associations between different variables within or between partners can be referred to as bivariate ICCs (Kenny et al., 2006). Interpretation of the ICCs was similar to that of the Pearson’s rs. As displayed in Table 3, IHP, psychological IPV perpetration, and physical IPV perpetration were negatively associated with same-sex couple relationship quality. IHP were positively associated with psychological IPV perpetration and physical IPV perpetration.

Associations among IHP, IPV, and Relationship Quality

Standardized coefficients for path analyses were displayed in Figure 1.

For H1, negative actor associations (b = −.26, S.E. = .09, p < .01, β = −.16) were identified between IHP and relationship quality. H1 was partially supported.

For H2, positive actor (b = .37, S.E. = .16, p < .05, β = .13) and partner (b =.43, S.E. = .16, p < .01, β = .15) associations were identified between IHP and psychological IPV perpetration. H2 was partially supported.

For H3, negative actor (b = −.12, S.E. = .04, p < .01, β = −.23) and partner associations (b = −.12, S.E. = .04, p < .01, β = −.21) were identified between psychological IPV perpetration and relationship quality. H3 was partially supported.

The Mediating Role of Intimate Partner Violence

For H4, four mediating pathways were identified (seen in Table 4). Individuals’ IHP was negatively associated with their own relationship quality via their own (b = −.046, 95% CI [−.100, −.004], β = −.029) and their partner’s psychological IPV perpetration (b = −.050, 95% CI [−.104, −.009], β = −.031). Individuals’ IHP was negatively associated with their partner’s relationship quality via their own (b = −.053, 95% CI [−.111, −.011], β = −.033) and their partner’s psychological IPV perpetration (b = −.043, 95% CI [−.095,.−.003], β = −.027). Based on Kenny’s (2012) criteria, all four identified indirect effects were between “small” and “medium” in terms of effect size. H4 was partially supported.

Table 4.

The specific indirect effects for each indirect pathway in the model based on bias-corrected bootstrap estimates (N = 144 couples)

Specific pathways tested in the model		Bootstrap estimates for indirect effects
		Unstandardized	95% CI	Standardized
Own IHP → Own relationship quality
Specific indirect pathways	via own psychological IPV perpetration	−.046	[−.100, −.004]	−.029
	via partner’s psychological IPV perpetration	−.050	[−.104, −.009]	−.031
	via own physical IPV perpetration	−.001	[−.024, .017]	−.001
	via partner’s physical IPV perpetration	−.007	[−.032, .007]	−.004
Own IHP → Partner’s relationship quality
Specific indirect pathways	via own psychological IPV perpetration	−.053	[−.111, −.011]	−.033
	via partner’s psychological IPV perpetration	−.043	[−.095, −.003]	−.027
	via own physical IPV perpetration	−.001	[−.017, .017]	−.001
	via partner’s physical IPV perpetration	−.112	[−.294, .044]	−.005

Note. IHP = Internalized Homophobia, IPV = Intimate Partner Violence. Bolded are indirect pathways that were significant at p < .05 (two-tailed). Physical IPV perpetration included in the table is the dichotomized score. IHP, psychological IPV perpetration, and relationship quality were transformed scores.

Additional Analyses

The cross-sectional nature of the present data did not allow the examination of the temporal ordering of study variables, although it is possible that the associations among IHP, IPV, and same-sex relationship well-being may be bidirectional (Balsam & Szymanski, 2005). As such, we tested two set of alternative models that were proposed in existing studies (Balsam & Szymanski, 2005). Covariates included in the primary analyses were also controlled for in the examinations of the alternative models.

In the Alternative Model 1, we conducted an APIMeM to examine whether same-sex relationship well-being mediated the associations between IHP and IPV. Comparative fit index (CFI) for the Alternative Model 1 is .02 lower than that for the model depicted in Figure 1. In the Alternative Model 2, we conducted another APIMeM to examine whether IPV will predict IHP via same-sex relationship well-being. CFI for the Alternative Model 2 is .14 lower than that for the model depicted in Figure 1. As a decrease in CFI by .01 or more indicates a significantly worse fitted model (Kline, 2015), we concluded that the model in Figure 1 was preferred than the Alternative Models 1 and 2.

Discussion

The present study complements and extends prior research in important ways. First, we revisited the association between IHP and same-sex relationship quality using couple dyadic data from a more geographically and socioeconomically diverse sample and also controlling for an extensive set of covariates. Second, guided by a process perspective and utilizing a more rigorous statistical strategy (i.e., the APIMeM with interchangeable dyads), this study was among the initial steps in examining the potential mediating roles of IPV perpetration in the association between IHP and same-sex relationship quality. Our findings contribute to the understanding of the mechanisms underlying the implications of IHP, as a more proximal sexual minority stressor, for same-sex relationship well-being and also provide unique insights to interventions targeted at assisting same-sex couples to navigate through the difficulties associated with their historically disenfranchised status.

The first noteworthy finding in the present study is about the characteristics and prevalence of psychological and physical IPV perpetration in the present sample. On one hand, the bidirectional IPV perpetration (i.e., both partners within a couple perpetrated IPV) in the present study are more common than unidirectional IPV perpetration (i.e., only one partner within a couple perpetrated IPV), which is in line with the previous findings based on community samples of both different-sex and same-sex population (for a review, see Edwards et al., 2015). One of the most pervasive explanations is that, as the two partners of couples in community sample typically has comparable power, one partner will “fight back” when the other partner perpetrated IPV (for a review, see Rollè, Giardina, Caldarera, Gerino, & Brustia, 2018). The explanation of “comparable power” may be particularly true for same-sex couples, as two partners in same-sex relationships were typically regarded as equal in terms of physical strength and social status due to their same-sex status (Rollè et al., 2018).

On the other hand, the 12-month prevalence of IPV perpetration in the present study is higher than the 12-month prevalence of IPV perpetration in prior studies based on community sample of general-population (e.g., Elliott et al., 2011; Straus, 2017) and same-sex population (e.g., Edwards & Sylaska, 2013; Finneran & Stephenson, 2013b). Two explanations may be applicable. First, the higher prevalence in comparison to the general populations highlights the necessity of identifying LGBTQ-specific variables as potential precursors for IPV in same-sex relationships (Baker, Buick, Kim, Moniz, & Nava, 2013; Edwards, Sylaska, & Neal, 2015; Mason, Lewis, Milletich, Kelley, Minifie, & Derlega, 2014; Murray & Mobley, 2009). As stated already, the historically disenfranchised status renders same-sex couples to experience an additional set of stressors (i.e., LGBTQ-specific minority stressors) that are not shared by different-sex couples (Frost, 2011). The additional set of stressors may then proliferate into the high levels of IPV perpetration in same-sex community. Second, the higher prevalence in comparison to same-sex samples in prior studies may be due to the socioeconomic diversity of the present sample. As noted already, existing studies in the field of same-sex couple relationships were based predominantly (or even exclusively) on middle-class, Non-Hispanic White couples. Our sample included a notable proportion of couples who were racial/ethnic minority and/or living in low-income status, which also might increase the risk of IPV perpetration (Edwards et al., 2015).

The central finding of the present study is that same-sex spouses’ IHP are associated with spouses’ own (i.e., the actor associations) and their partners’ relationship quality (i.e., the partner’s association) quality via spouse’ own and their partners’ psychological IPV perpetration. To begin with, our finding adds to a limited body of evidence supporting the dyadic implications of IHP for same-sex relationship well-being (Feinstein et al., 2018; Otis et al., 2006). Given the interdependence between the two partners in a same-sex relationship, individuals’ IHP (i.e., an intrapersonal stressor) may not only spillover to their relational domain and diminish their own relationship well-being, but also crossover to compromise their partners’ relational happiness (Otis et al., 2006). From a methodological perspective, our finding highlights the importance of collecting data from both partners in a same-sex couple and employing dyadic data analytic strategies such as the APIM and its derivatives (e.g., APIMeM and APIMoM) to more adequately tackle the complexity inherent within the implications of sexual minority stressors for same-sex relationship well-being (Garcia, Kenny, & Ledermann, 2015; Kenny et al., 2006; Ledermann et al., 2011; Smith, Sayer, & Goldberg, 2013).

More importantly, the identification of the mediating roles of spouse’ own and their partners’ psychological IPV perpetration not only lends empirical support to the classic proposition in the VSA model (i.e., vulnerabilities → adaptive process → relationship outcome; Bradbury & Karney, 1995), but also provide practical implications for prevention and intervention programs that aimed at reducing IPV and improving same-sex relationship well-being (Edwards et al., 2015). That is, practitioners should pay special attention to same-sex couples in which one or both partners having high levels of IHP when working with same-sex clients troubled by IPV and relationship dissatisfaction (Carvalho, Lewis, Derlega, Winstead, & Viggiano, 2011; Finneran et al., 2012; Lewis et al., 2014). As noted already, high levels of IHP were related to (a) negative attitudes and expectations toward their same-sex relationships, (b) avoidance of intimacy in same-sex relationships, and (c) reduced energy to take care of their partners’ needs and maintain relationship (LeBlanc et al., 2015; Meyer, 2003; Meyer & Dean, 1998; Otis et al., 2006). For those with high levels of IHP, the negative attitudes for same-sex relationships may increase the likelihood for their own IPV perpetration (Balsam & Szymanski, 2005; Finneran et al., 2012; Lewis et al., 2014). For partners of those with high levels of IHP, the avoidance of intimacy and the decreased relationship maintenance may be regarded as a signal of withdrawal and indifference (Stanley, Bartholomew, Taylor, Oram, & Landolt, 2006). To get greater closeness and satisfy emotional needs, partners of those with high levels of IHP may utilize IPV perpetration as the last resort to keep the avoidant spouses involved in couple interactions (Stanley et al., 2006). Considering that IPV perpetration is associated with negative feelings of both the perpetrators (e.g., shame and guilt; Stith, Green, Smith, & Ward, 2008) and the victims (e.g. fears and loss of power; Caldwell et al., 2012), IPV perpetrated by one partner in turn diminish both spouses’ relationship quality.

Notably, whereas psychological IPV perpetration plays salient roles in the associations between IHP and relationship well-being, no significant pathways involving physical IPV perpetration were identified. Several explanations may help understand such salient roles of psychological IPV perpetration. First, psychological IPV involves a series of behaviors that are harmful for the victims as an individual and the couple as a relationship unit. For instance, prior studies suggest that psychological IPV is a stronger predictor of victims’ fear than physical IPV, presumably because the perpetrator’s threatening behavior (e.g., threatening to hit the other, which is an indicator for psychological IPV) engender prolonged stresses on the victim (for a review, see O’Leary, 1999). Such negative feelings in turn diminish couple relationship well-being and increase victims’ tendency to end the relationship (Arias & Pape, 1999). Further, the perpetrator’s refusing to talk (i.e., another indicator for psychological IPV) may thwart not only victims’ attempts to seek intimacy but also the solution of specific relationship problems (Panuzio & DiLillo, 2010). Unresolved relationship problems will linger and sow the seeds for relationship distress in the future (Christensen & Heavey, 1990). Second, psychological IPV is relatively common in couple relationships (Mason et al., 2014). As psychological IPV reoccurs, the negative consequences of psychological IPV are likely to accumulate and explode (Schumacher & Leonard, 2005). Third, the relatively low prevalence of physical IPV in the present sample may limit the statistical power to detect significant role of physical IPV perpetration in the associations between IHP and same-sex relationship quality.

Limitations and Future Directions

Some limitations of the present study should be noted. First, data in the present study were collected from a community sample, and participants in the present study may report higher relationship quality whereas less psychological and physical IPV perpetration than couples experiencing severe relational problems (e.g., couples in abusive relationships; Edwards et al., 2015). Moreover, we only included couples in which both partners participated in the larger project. Based on the attrition analyses, participants included in the present sample perpetrated less physical IPV than those from couples in which only one partner participated in the larger project. Thus, the prevalence of IPV perpetration (especially physical IPV perpetration) in the present sample may be underestimated. Generalizability of our findings also should be made with caution.

Second, the sample size of the present is modest, although it is comparable to those of recent studies (e.g., Feinstein et al., 2018). This may limit the statistical power in analyses. Notably, recruiting sexual minority partners, especially both partners in a same-sex couple, is challenging. Some sexual minority partners may withdraw given the concerns that participation is associated with “coming out” and may incur discrimination (Meyer & Wilson, 2009). That may also be the reason why IHP in the present study were at the lower end of the used scale (i.e., the mean is 2.02 out of 7). Replications of the present findings with larger and more representative samples are warranted.

Third, the cross-sectional and correlational nature of the present analyses prevents us from clarifying the directionality of the examined associations and from reaching any causal conclusions. Although the data in the present study supported our proposed model (i.e., IHP → IPV → same-sex relationship well-being) rather than the alternative models (as presented in the additional set of analyses), longitudinal studies are still pressing.

Fourth, given that all key variables in the present study were assessed via self-report questionnaires, the currently identified associations might have been inflated by the shared-informant and shared-method variance. Also, considering the potential social desirability bias in self-report surveys, the levels of IHP and IPV perpetration might be underreported, whereas the levels of same-sex couple relationship quality might be overreported in the present study. To more adequately capture the study constructs and the associations among them, future studies would benefit from utilizing multi-method and multi-informant designs.

Last, IHP is often viewed as an individual-level (or intrapersonal) stressor, and small-sized correlation emerged between two partners’ IHP in the present study, which suggest the potential discrepancies in IHP between the two partners in a same-sex couple. Further, IHP also may be experienced jointly by the couple as a result of the stigmatized status of their same-sex relationships (e.g., “Believing that one’s same-sex relationship is less valuable to society than heterosexual relationships are”). Examinations of the unique and joint implications of individual-level IHP, the discrepancies in IHP between partners in a couple, and couple-level IHP for same-sex relationship well-being are promising (see LeBlanc et al., 2015 for a detailed discussion).