Abstract
Intimate partner violence (IPV) is a major public health concern, leading to prevention efforts focused on identifying risk indicators of escalating conflict between partners. Certain behaviors during conflict discussions have been uniquely linked to IPV, and there is evidence that disengagement behaviors—an emerging construct—are associated with IPV as well. However, research on disengagement and IPV has largely been limited to self-report, and available observational research is largely cross-sectional. Addressing these limitations, this study prospectively examined the association between observed disengagement behaviors during couple’s conflict discussions and IPV 1 year later, using a sample of 83 heterosexual married or cohabitating partners. Behavioral coding was used to assess each partner’s observed disengagement behaviors and each partner’s psychological and physical IPV was assessed via questionnaire 1 year later. Linear regressions were used to investigate links between indicators of engagement (i.e., listening, questions, eye contact) and psychological and physical IPV perpetration and victimization. Less question asking and higher eye contact by women were associated with greater psychological and physical IPV perpetrated by both women and men 1 year later. Lower listening behaviors by women were associated with greater psychological (but not physical) IPV perpetrated by men 1 year later. In contrast, no disengagement behaviors displayed by men were significantly related to IPV. Applying a latent change score framework using baseline IPV data, findings indicated replicability of results, though with less consistency, suggesting that women’s disengagement behaviors may reflect a larger pattern of abuse that predates and follows disengaged couple interactions. Findings suggest that a unique blend of verbal and nonverbal indicators of women’s disengagement during couple conflict provides a meaningful signal of the emotional climate of the relationship.
Keywords
Intimate partner violence (IPV) is a major public health concern in the United States. Each year, one in three women and one in four men experience physical IPV, which has been linked with physical, mental, sexual, and reproductive health effects (National Coalition Against Domestic Violence, 2020). Identifying risk indicators of IPV is a critical step in developing interventions to prevent the onset of violence, reducing violence once it is present, and mitigating the health consequences of IPV. One group of risk indicators with the potential to contribute to conflict are the behaviors that couples display during conversations about points of disagreement. Certain kinds of behaviors that couples engage in when discussing contentious topics are associated with IPV, including demand–withdrawal patterns (Babcock et al., 1993; Holtzworth-Munroe et al., 1998; Pickover et al., 2017), ending positive engagement (Ha et al., 2019), and negative conflict resolution behaviors (Jarnecke et al., 2020). Some evidence suggests that disengagement—an emerging construct—is similarly associated with IPV (Singh, 2018). However, though disengagement has been linked to concurrent IPV when measured via questionnaire (Singh, 2018), it is unknown whether observed indicators of disengagement are associated with future IPV behaviors.
Demand–Withdrawal Communication Patterns
The demand–withdrawal dyadic communication pattern has been a focused area of couple communication research for decades (Christensen & Heavey, 1990; Crenshaw et al., 2021; Eldridge & Baucom, 2012). This pattern describes an interaction between couples in which one partner demands, criticizes, or pressures in a conversation while the other withdraws, avoids confrontation, or changes the topic (Christensen, 1988). Meta-analytic findings suggest that demand–withdrawal relationship patterns are significant risk factors of physical IPV perpetration across men and women, with the strongest effects emerging among men (Spencer et al., 2020). In other words, stronger demand–withdrawal dynamics in which one partner is in the “demanding” role while the other takes a “withdrawing” role is associated with an increased risk of physical IPV, and this risk factor is significantly stronger among men than among women. Similarly, a recent study of IPV-exposed women found that psychological IPV specifically was associated with partner-demand/self-withdraw, again suggesting that women in violent relationships may withdraw to escape conflict and secure safety (Pickover et al., 2017). In contrast, other research has found that physically violent (compared to nonviolent) husbands were more likely to demonstrate both demanding and withdrawing behaviors during observed couple interactions while wives were more likely to demonstrate demanding behaviors but were less likely to withdraw, possibly out of a desire to avoid further demanding from violent husbands (Berns et al., 1999). Clearly, more research is needed to better understand the complex links between IPV and couple communication dynamics.
Couple Disengagement
The withdrawal component of the demand–withdrawal cycle has conceptual overlap with a related but unique construct known as disengagement. Definitions of disengagement have varied, in part because the concept is characterized by the absence (as opposed to the presence) of negative or positive couple behaviors (e.g., lack of interest, lack of disclosure; see Barry et al., 2008 for a discussion). In an attempt to synthesize existing indicators of disengagement, Barry and colleagues used exploratory factor analysis to develop and validate the Romantic Disengagement Scale (RDS; Barry et al., 2008). Findings indicated that disengagement is comprised of a unidimensional factor that include behavioral distancing (e.g., avoidance, speaking less), emotional indifference (e.g., feeling apathetic), and cognitive distancing (e.g., distracting oneself; Barry et al., 2008). Emotional indifference is typified by low levels of positive emotion coupled with low levels of negative emotion (Gottman, 1999; Kayser, 1993, 1996), while cognitive and behavioral strategies that create psychological or physical distance between two partners include behaviors such as avoidance, withdrawing, being less attentive toward one’s partner, and refraining from personal disclosure with one’s partner (Gottman, 1999; Kayser, 1993, 1996). Using the RDS, Singh (2018) found that self-report disengagement was positively associated with one’s own and one’s partner’s concurrent (though not future) physical IPV among newlyweds.
Given evidence of discrepancies between individuals’ self-reports of communication and reports made by their spouses (e.g., Rhoades & Stocker, 2006), there is a need to extend prior research examining self-report disengagement to investigate whether observed couple behaviors predict IPV as well. Notably, observational studies have linked behaviors that may overlap with disengagement with concurrent or future IPV (e.g., Ha et al., 2019; Jarnecke et al., 2020; Sommer et al., 2019); however, these studies’ observational paradigms do not fully capture the construct. For example, Ha et al. (2019) examined negative (e.g., stonewalling) and positive (e.g., active listening) interaction patterns but utilized scores to develop a measure of dyadic coercion as opposed to disengaged communication. Sommer et al. (2019) examined eyerolling, insults, hostile humor, mockery, facial disgust, sarcasm, disrespectful comments, and name-calling, all of which are more akin to contempt. Jarnecke et al. (2020) examined nine types of conflict behavior (including withdrawal) to draw associations between IPV and positive and negative conflict resolution strategies. Still, further observational research drawing from multifaceted assessment of disengagement (i.e., based on factor-analyzed models) is needed. The purpose of this study is to examine whether observed disengagement behaviors during couple’s conflict discussions predict physical and psychological IPV 1 year later.
The Present Study
The above findings suggest that disengagement during partner conflict might serve as behavioral risk indicators of physical and psychological IPV. While observational research has suggested that elements of disengagement are related to IPV, no study to date has examined the association between disengagement (defined by observable behaviors reflecting emotional indifference, behavioral distancing, and cognitive distancing) and IPV. There is also a need for more longitudinal research on this topic, as much of the research has been cross-sectional (e.g., Jarnecke et al., 2020; Pickover et al., 2017). Consequently, we seek to understand whether disengagement during couple’s conflict discussions predicts physical and psychological IPV a year later. Based on prior work investigating similar constructs (e.g., demand–withdrawal patterns; Pickover et al., 2017), we expected that disengagement during conflict would predict psychological and physical IPV 1 year later.
Importantly, if disengagement during conflict predicts future IPV, this may reflect an attempt by victims to protect themselves given that disengaged behaviors may initially arise in the context of ongoing violence. As such, disengagement behaviors may reflect behavioral indicators of risk that emerge as part of a larger pattern of violence that predates and follows instances of IPV. Indeed, there is evidence to suggest that physical IPV predicts future self-reported disengagement (Singh, 2018). Put differently, behavioral and emotional distancing behavior reflective of disengagement likely does not emerge in a vacuum, but rather, transpires as natural learned reactions to psychological or physically violent behavior. In this way, we conceptualize disengagement as an indicator of the emotional climate of the relationship, as opposed to the cause of violence.
Method
Participants
Participants were drawn from a 1-year longitudinal investigation of post-traumatic stress disorder, relationship abuse, and physical health among couples (see Gilbar et al., 2021; LaMotte et al., 2016). Recruitment occurred within the greater Boston area between 2008 and 2011 via advertisements soliciting couples to participate in a study on trauma and relationship problems. The sampling strategy oversampled for couples experiencing IPV by assessing whether any of the couple’s disagreements had become physical in the past 6 months on a phone screen. However, to facilitate variability, a sizeable minority of participants who responded negatively to this question were also included in the sample. Participants were recruited via four sources: (a) online postings on job boards at a local university, (b) online postings for volunteers on Craigslist, (c) flyers posted in public areas, and (d) newspaper advertisements in local papers. Heterosexual couples between the ages of 18 and 60, who reported being married or cohabitating for at least 6 months, and able to read and speak English were included in this study. Inclusion criteria also stipulated that both members of the couple agree to participate. Participants were excluded if they were pregnant or currently trying to become pregnant, or if they reported being in inpatient treatment for substance use problems or beginning methadone maintenance in the last 2 months.
The final sample consisted of 83 heterosexual couples (166 individuals). Participants were typically in their mid-30s. About one-third of the sample was married and two-thirds of the sample were unmarried and living with their partners. Couples reported being in their relationships an average of 6.5 years. Couples reported an average of 14 years of education; total annual gross income (not including partner’s) was $10–20k. Approximately half the sample identified as White. See Table 1 for additional demographic information.
Sample Characteristics (N = 83 Dyads).
Procedures
All procedures were approved by the VA Boston Healthcare System and Boston University Medical Center review boards. Participants were screened over the phone to determine eligibility. Eligible participants were scheduled for two baseline sessions spaced 1 week apart to reduce biased responses to assessments (Groves et al., 2009). Upon arrival to their first session, participants provided written informed consent, which included agreement to be re-contacted at 6- and 12-month follow-ups. Each member of the couple underwent identical procedures and separately completed questionnaires to ensure privacy. During the second baseline assessment, participants completed laboratory procedures, including the video-recorded task used in this study. The IPV measure used in this study was administered at baseline and 1 year later to assess IPV since baseline. Participants received $40 for completing the first baseline assessment, $60 for completing the second baseline assessment, $30 for completing the 6-month assessment, and $30 for completing the 1-year assessment.
Measures
The Revised Conflict Tactics Scale-2 (CTS2; Straus et al., 1996) was used to assess psychological and physical IPV perpetration and victimization. Psychological IPV was assessed with the eight-item Psychological Aggression subscale (e.g., I insulted or swore at my partner) and Physical IPV was assessed with the 12-item Physical Assault subscale (e.g., I pushed or shoved my partner). The CTS-2 is widely used self-report measure of IPV and has evidence of construct validity (Straus et al., 1996), as well as excellent test–retest reliability (Vega & O’Leary, 2007), and acceptable to excellent internal consistency (Straus et al., 1996). Participants reported the frequency with which they and their partners engaged in each behavior over the past 6 months. As is the gold-standard approach, to reduce underreporting, we compared both partners’ reports on each item and used the higher of the two responses prior to creating a sum score. For the assessment of physical aggression, we used variety scores (i.e., dichotomous endorsement of each item), which we subsequently square root transformed; this enabled us to reduce skewness created by a small number of participants with high frequencies of physical aggression. In contrast, we used frequency scores for the assessment of psychological aggression, which is more appropriate given higher occurrences of psychological aggression in the sample.
To assess romantic disengagement, study staff instructed each partner to identify a problem that was a source of tension in the relationship using the Marital Problem Inventory (Geiss & O’Leary, 1981). Next, couples were asked to discuss the relationship problem selected by each member in a randomized order. In situations when partners chose the same topic for discussion, the spouse selected to be first was assigned that topic; the other partner was encouraged to discuss their second topic during the second discussion. Couples were instructed to “discuss the topic for 10 minutes and try to work toward a mutually satisfying solution.”
A modified version of the Romantic Disengagement Behavioral Coding system (Barry, 2010) was used to code each partner’s behavioral disengagement during the video-recorded conflict interaction. The original measure included 17 codes capturing important components of disengagement suggested in previous research (Hess, 2002; Heyman & Vivan, 2000; Parkinson & Totterdell, 1999; Smith et al., 1990); however, based on the author’s previous use of the measure (e.g., Barry, 2010), these 17 codes were reduced to six codes for this study. These six codes were chosen because they demonstrated higher reliability and validity in the original development of the coding system; furthermore, scores on these codes yielded greater variability (i.e., certain codes rarely observed were dropped; Barry, 2010). These six codes included (a) avoiding the discussion (e.g., refusing to talk about an issue), (b) shutting out the partner behaviorally (e.g., refusing to make eye contact), (c) showing low evidence of listening to the partner (e.g., showing few/no verbal and nonverbal behaviors occurring in response to a speaker’s words), (d) brushing off the partner (e.g., verbally expressing denial, disinterest, or giving up), (e) showing low evidence of interest in the discussion (e.g., asking few/no questions that express genuine interest), and (f) low levels of verbal engagement (e.g., less frequently speaking and giving appropriate time for partner to respond). After coding procedures were complete, analysis of descriptive statistics revealed severely restricted ranges of some of the disengaged behaviors (i.e., floor effects). This necessitated that three codes be dropped: avoiding the discussion, brushing off, and speaking. The remaining codes (i.e., listening, questions, eye contact) were used in this study and were coded separately for each member of the couple. Higher scores are indicative of higher levels of disengagement (i.e., lower evidence of listening, less questions asked, and fewer seconds of eye contact).
Trained research assistants coded the presence of each behavior for each partner for each 10-second time segment. Listening behaviors were coded on a scale of 1–5 for each segment, with lower scores indicating frequent behavioral evidence that the target is listening closely, moderate scores indicating evidence of listening behavior for about half the segment, and high scores indicating no evidence that the target is listening. Listening behaviors are coded in response to the speakers’ words and include both short verbal behaviors (e.g., “yeah,” “right,” “I see”), verbal tracking (i.e., replies that suggest the target actually heard the words the partner said), as well as nonverbal behaviors (e.g., nodding, shaking their head, shrugging, raising eyebrows, smiling in response to something the partner said). Questions were coded on a scale of 1–3 for each segment, with lower scores indicating that at least one question was asked expressing genuine interest and higher scores indicating that no questions were asked. Eye contact was coded by counting the total number of seconds out of 20 seconds in which the target was looking at their partner’s face, attempting to make eye contact. A total score for each code was computed by averaging across intervals. Coders trained for approximately 10 hours to reliability or longer if needed to achieve strong agreement. Inter-rater reliability was established by double coding a random sample of 22% of the interactions and indicated moderate to strong inter-coder agreement (listening = .61, questions = .80, eye contact = .88; Koo & Li, 2016).
Data Analytic Plan
Data were screened for normality and patterns of missing data. Outcome variables were normally distributed based on accepted values for skew and kurtosis (Kline, 2015). For primary analyses, separate linear regressions were used to examine associations between indicators of disengagement (i.e., listening, questions, eye contact), entered simultaneously, and continuous IPV variables, separately among men and women. This resulted in eight regression analyses (i.e., men’s disengagement behaviors predicting physical and psychological IPV perpetration and victimization, and women’s disengagement behaviors predicting physical and psychological IPV perpetration and victimization). Consistent with a parsimonious approach (Becker et al., 2016), demographic variables associated with both the predictor variables (i.e., disengagement behaviors) and outcome variables (i.e., physical and/or psychological IPV) were controlled for in regressions. Next, to examine whether disengagement behaviors predicted within-person change in IPV (i.e., the degree of change over 1 year), we applied a latent change score framework (Castro-Schilo & Grimm, 2018; McArdle & Nesselroade, 1994). The latent change score approach, in contrast to the computation of raw change scores, does not drop cases with missing follow-up scores. The models were just identified and thus global fit was not assessed.
Results
Preliminary Analyses
A sizeable amount of data was missing on primary predictor variables (i.e., video-coded behaviors, ranging from 18% to 23% of the sample across codes); these data were missing due to corrupted video files and were assumed missing at random. Furthermore, 17% of participants did not participate at the 1-year follow-up session, resulting in data missing on the primary outcome variables (i.e., psychological and physical IPV). In total, 41% of participants were missing data on at least one variable. To determine whether these data were missing at random, we examined bivariate correlations between demographic variables and missingness. Results indicated that men with missing IPV data were more likely at baseline to have reported lower incomes and more likely to identify as Black; both men and women with missing IPV data were more likely to be younger, report shorter relationships, and report living together for less time. Consequently, data were assumed to be not missing at random and were addressed using multiple imputation (i.e., entering in demographic data as auxiliary variables).
Demographic variables (i.e., age, marital status, relationship length, length of time living together, years of education, individual gross income) were considered as possible covariates in primary analyses; results indicated that only age was associated with disengagement behaviors (i.e., fewer questions asked; r = .34, p < .01; less eye contact; r = .28, p < .05) and men’s psychological IPV perpetration (r = .34, p < .01) and was subsequently controlled for in primary analyses. Given that age was included as a covariate in analyses, we only included age as an auxiliary variable in multiple imputation for the opposite gender (i.e., analyses examining the effects of men’s disengagement behaviors only included women’s age as an auxiliary variable).
Descriptive information about primary variables of interest are presented in Table 1. Regarding frequencies, nearly all couples reported at least one incident of psychological aggression perpetration at baseline (97.6%) and at follow-up (92.0%). In contrast, physical aggression perpetration was reported by just over half of couples at baseline (56.7%) and just over a third at follow-up (37.7%). Examination of mean rates of perpetration and victimization among men and women suggests that on average, women and men engaged in similar levels of psychological and physical IPV toward each other at both baseline and 1-year follow-up. The only exception was physical IPV at baseline, which women perpetrated at significantly greater rates compared to men, t(82) = −3.06, p = .003. Bivariate correlations among primary variables of interest are presented in Supplemental Table 1. Overall, physical and psychological IPV perpetration among partners were highly correlated, such that greater aggression was significantly associated with greater aggression in one’s partner. Moderate to large effect sizes also emerged between disengagement behaviors among partners, such that poorer listening, fewer questions, and less eye contact behaviors were associated with greater disengaged behaviors in one’s partner. Small to medium correlations between the three coded disengagement behaviors emerged among both men and women but were sufficiently distinct (rs < .70) to examine separately.
Associations of Women’s and Men’s Disengagement Behaviors with IPV Perpetration by Women 1 Year Later
Results of each regression model are presented in Table 2. After adjusting for women’s age, less question asking by women (B = 74.25, 95% CI [45.34, 103.15], p < .001) and higher eye contact by women (B = −1.67, 95% CI [−2.26, −1.08], p < .001) were associated with greater psychological IPV perpetrated by women 1 year later. The model explained 55.9% (p < .001) of the variance in the outcome. Similarly, less question asking by women (B = 1.40, 95% CI [0.22, 2.53], p < .05) and higher eye contact by women (B = −0.03, 95% CI [−0.06, −0.01], p < .01) were associated with greater physical IPV perpetrated by women 1 year later. The model explained 25.4% (p < .05) of the variance in the outcome. In contrast, after adjusting for men’s age, no disengagement behaviors displayed by men were significantly related to women’s IPV perpetration 1 year later.
Results of Regression Models (N = 83 Dyads).
Note: All behavioral observations are coded such that higher scores indicate greater levels of disengagement (i.e., less listening, fewer questions, lower eye contact). B = unstandardized estimates; β = standardized estimates; CI = confidence interval; IPV = intimate partner violence; SE = standard error.
p Values are reported for unstandardized estimates. Significant effects (p < .05) are bolded.
Associations of Women’s and Men’s Disengagement Behaviors with IPV Perpetration by Men 1 Year Later
After adjusting for women’s age, fewer listening behaviors displayed by women (B = 15.01, 95% CI [1.47, 28.54], p < .05), less question asking by women (B = 77.10, 95% CI [45.55, 108.65], p < .001), and higher eye contact by women (B = −1.76, 95% CI [−2.39, −1.13], p < .001) was associated with greater psychological IPV perpetrated by men 1 year later. The model explained 55.5% (p < .001) of the variance in the outcome. Similarly, less question asking by women (B = 1.37, 95% CI [0.08, 2.72], p < .05) and higher eye contact by women (B = −0.03, 95% CI [−0.06, −0.01], p < .01) were associated with greater physical IPV perpetrated by men 1 year later. The model explained 27.8% (p < .05) of the variance in the outcome. In contrast, after adjusting for men’s age, no disengagement behaviors displayed by men were significantly related to men’s IPV perpetration 1 year later.
Latent Change Framework
Results for disengagement behaviors predicting IPV latent change scores are reported in Supplemental Table 2. Regarding men’s IPV perpetration, results remained fairly consistent. Less question asking by women (B = 0.96, SE = 0.47, p = .040) and greater eye contact (B = −0.02, SE = 0.01, p = .025) were associated with greater increases in men’s physical IPV perpetration. Similarly, less question asking by women (B = 45.40, SE = 17.46, p = .009) was associated with greater increases in men’s psychological IPV perpetration. The relationship between greater eye contact among women and men’s psychological perpetration dropped to marginally significant (B = −0.76, SE = 0.386, p = .051) and women’s listening behaviors no longer significantly predicted an increase in men’s psychological IPV perpetration. Regarding women’s IPV perpetration, results changed more considerably. Women’s disengagement behaviors were no longer associated with greater increases in women’s physical or psychological IPV perpetration over the subsequent year, with the exception of less question asking by women (B = 43.54, SE = 19.84, p = .028), which continued to predict greater increases in women’s psychological IPV perpetration. Finally, analogous to the primary results, men’s disengagement behaviors predicted neither women’s nor men’s IPV perpetration.
Discussion
The purpose of this study was to examine the role of observed disengagement—an understudied construct reflecting behavioral and cognitive distancing, as well as emotional indifference—in predicting physical and psychological IPV between couples 1 year later. Findings of main analyses revealed that less question asking and higher eye contact by women were associated with greater psychological and physical IPV perpetrated by women 1 year later. Less question asking and higher eye contact by women were also associated with greater physical IPV perpetrated by men 1 year later, and the same behaviors, in addition to fewer listening behaviors, were associated with greater psychological IPV perpetrated by men 1 year later. In contrast, no disengagement behaviors displayed by men were significantly related to IPV. Findings of supplementary analyses using a latent change framework indicated that some of these associations were no longer significant when accounting for baseline levels of IPV. A discussion of overall results, followed by exploration of findings relevant to specific disengagement behaviors, follows.
Prior research using self-report data demonstrated that greater disengagement predicted greater concurrent use of IPV by both individuals and their partners at baseline, but not at follow-up (Singh, 2018). In comparison, we demonstrated that women’s disengaged behaviors predicted men’s and women’s IPV perpetration behaviors 1 year later to some extent; however, a number of findings in this study were either null (i.e., particularly regarding men’s disengagement behaviors) or in the opposite direction (i.e., greater eye contact predicted greater IPV), particularly when taking into account latent difference scores. Indeed, our finding that some of the associations between women’s disengaged behavior and IPV 1 year later disappeared when accounting for baseline IPV is poignant in that it lends credibility to the notion that women’s disengagement behaviors do not indeed cause IPV. Rather, this aligns with a dynamical systems lens in which preexisting violence reinforces disengagement: women in relationships characterized by violence may increasingly learn that behavioral and emotional distancing facilitates protection from violent encounters. As such, disengagement may reflect an indicator of the potential for violence. For example, women who believe they are at risk might pay closer attention to gauge the temperature of the room (i.e., facilitating greater eye contact), while being careful not to say something that might anger their partner (i.e., facilitating less question asking). Taken together, the different effects emerging from our primary and supplementary analyses reflect that disengaged communication patterns occurring in the context of conflict likely emerge as part of an ongoing pattern that gives rise to behaviors aimed at de-escalating the situation.
Gender Effects
In our sample of couples, women’s—but not men’s—disengagement behaviors appeared to indicate risk for future IPV perpetration by both men and women. Findings suggest that a unique blend of verbal (i.e., listening, question asking) and nonverbal (i.e., eye contact) indicators of women’s disengagement during couple conflict provides a meaningful signal of the emotional climate of the relationship. It is notable that women’s disengagement behaviors were associated with greater perpetration of IPV by women, as well as more experiences of victimization by women. This finding may have emerged because the sample was largely characterized by mutual bidirectional violence (i.e., members of a couple reported perpetrating violence at similar rates); if women are perpetrating IPV in response to their partner’s IPV, this suggests that women’s disengaged communication may reflect preemptive attempts to avoid engaging in contentious discussions that could ultimately result in women needing to defend themselves by responding to violence with reciprocal attacks.
Yet, why would women’s—but not men’s—disengagement behaviors predict IPV perpetration and victimization? One possibility is that women are more emotionally expressive than men, across both positive and negative emotions (Kring & Gordon, 1998; for reviews, see Ashmore, 1990; Brody & Hall, 1993). Though gender differences are rather small, meta-analytic results demonstrate that women are more likely to display hostility, distress, and intimacy during conflict compared to men, and these behaviors are associated with relationship satisfaction in the range of small to medium effects (Woodin, 2011). Furthermore, another meta-analysis showed that stronger conflict resolution skills are a significantly stronger predictor of lower IPV perpetration among women compared to men (Spencer et al., 2020). Taken together, it is possible that women’s behaviors are more predictive of relationship functioning, and IPV in particular. However, further research is needed to explore the mechanisms behind the gender differences noted in our findings.
Question Asking
In this study, we found that women who asked more questions during conflict discussions were less likely to report perpetrating or experiencing IPV from their partners 1 year later. Findings align with research showing that less negative conflict resolution behaviors and more positive communication skills are associated with lower IPV perpetration among both men and women (Spencer et al., 2020). However, it is important to consider the context in which question asking may be diminished in a relationship characterized by IPV. Questions may signal a sign of openness and lower anxiety more typical of a relationship in which it is safe to ask questions. Discussing the issues of conflict necessitate challenging interactions, that, when navigated ineffectively, can result in escalating negative interaction patterns that could set the stage for relationship aggression. Intentionally asking fewer questions may develop in the context of a relationship marked by violence because asking more questions might increase the likelihood of conflict. In contrast, in healthy relationships where it is safe to ask questions, querying one another in the context of a disagreement may signal to the other person a greater level of interest in understanding their perspective (Chen et al., 2010) and opens the door for increased positive communication (e.g., opportunities to listen, understand, and validate; Laurenceau et al., 1998).
Eye Contact
Contrary to expectation, greater eye contact by women was associated with higher IPV perpetration and victimization among women 1 year later. Eye contact is typically considered a marker of engaged communication (e.g., Burgoon et al., 1984), particularly in the context of affectionate interactions (e.g., Dadds et al., 2012). Eye contact signals active listening, which enables the other person to feel heard and validated, reinforcing healthy communication patterns. Yet in the context of victimization, greater eye contact may reflect anxious attempts to scan one’s partner for danger to ascertain potential threat to oneself. Alternatively, in the context of perpetration, greater eye contact often signals threat and directly precedes physical aggression among non-human primates (see Harrod et al., 2020 for a review). Some research has found this pattern among humans as well; for example, high eye contact and close proximity—when unaccompanied by smiling—may be perceived as communicating greater dominance and control (Burgoon et al., 1984).
Listening
Finally, fewer listening behaviors by women was associated with greater psychological (but not physical) IPV perpetrated by men 1 year later. In contrast, listening behaviors by men were not associated with either type of IPV by men or women. Lower listening behaviors by women may predict IPV victimization in that it may reflect general disengagement from a relationship typified by psychological violence. In this way, it is possible that women whose partners perpetrate psychological violence displayed fewer listening behaviors due to a history of incurring negative backlash.
There are multiple possible explanations for why listening behaviors were not predictive of the majority of examined outcomes. In this study, short verbal behaviors (e.g., “I see”) and nonverbal behaviors (e.g., nodding) were considered indicators of listening; however, while listening behaviors may reflect high-quality empathetic responses, they could also reflect negatively valanced defensive listening. In addition, couples may have used verbal and nonverbal indicators of listening without truly engaging in active listening. Active listening—in which a listener attempts to understand the speaker’s lived experience (Gordon, 1975; Rogers, 1951)—may be needed to facilitate the kinds of reflections and insights critical to navigating conflict effectively. Importantly, though listening behaviors are technically unobservable, the perception of listening is likely of high importance to couples in the context of salient conversations. Though the construct is challenging to capture in observational research (i.e., intercoder reliability of the listening code was moderate in this study), our coding system nonetheless enabled us to capture listening behaviors likely to be important to couples in the real world (e.g., nonverbal affirmations such as smiling or nodding in response to one’s partner). Still, future research could consider alternate methods for measuring listening behaviors, such as the degree to which partners perceive one another to be listening (e.g., Weger et al., 2010).
Limitations
This study had several limitations worth consideration. Although this prospective investigation represents an extension of prior cross-sectional work, we did not examine change in disengagement over time; thus, we are unable to draw conclusions about the extent to which change in disengagement predicts increases or decreases in IPV. Second, though our sample was fairly diverse in some regard (e.g., half the participants identified as racial and/or ethnic minorities), the sample consisted of only heterosexual couples. Interestingly, there is evidence to suggest that demand and withdrawal behaviors emerge similarly across same- and cross-sex couples (Baucom et al., 2010). Third, though we were able to examine three observable indicators of disengagement, some codes originally intended for analyses were dropped after observing floor effects that resulted in restricted ranges (i.e., avoiding discussion, brush-offs, and speaking). Future research may have more success in measuring these constructs among more severely violent couples, as well as among samples more diverse in conversational style. Fourth, the intercoder reliability was moderate only for listening behavior. Regardless, we still found women’s listening behavior to be significantly related to men’s psychological IPV perpetration, suggesting that reliability was high enough to capture this effect of interest. Lastly, there was some missing data for a considerable minority of the sample. Because participants with incomplete data may have experienced greater hardships and relational instability, this could have resulted in biased estimates. However, we addressed this using multiple imputation, which is the gold standard approach to deal with data that is not missing at random (Deng et al., 2016).
Clinical Implications
Results of this research have implications for intervention development. Given that women’s disengagement predicted both IPV perpetration and victimization, this suggests that these behaviors are indicative of a troubled relationship. However, it does not necessarily follow that disengagement behaviors should be targeted in treatment given that these behaviors may have arisen out of a need to minimize the likelihood of triggering verbal or physical attacks. Practitioners who work with couples need to screen for IPV and conduct a risk assessment to determine whether additional supports and safety planning is needed prior to considering whether it is safe to work with couples on their communication patterns, particularly because disengaged behaviors may even be protective in these situations. Treatment targeting communication will be most valuable for couples who wish to stay together and for those in which it is safe (i.e., distressed couples with lower intensity bidirectional IPV). In situations in which IPV is frequent and severe and risk factors for intimate partner homicide are present, targeting communication skills is likely not the immediate priority and may even result in iatrogenic consequences. In contrast, working on communication skills might be the appropriate strategy in situations where IPV is infrequent and couples are not at high risk for physical violence. Because different types of perpetrators have different treatment needs (Holtzworth-Munroe & Meehan, 2004; Messing et al., 2015), future research should investigate whether efforts to change patterns of communication engagement are more appropriate depending on the typology of IPV perpetration.
Notably, existing empirically supported treatment programs already target communication behaviors overlapping with disengagement. For example, Integrative Behavioral Couples Therapy, a therapy for distressed couples, encourages mutual engagement and emotional acceptance in the face of conflict (Doss et al., 2002). In addition, treatments aimed at perpetrators of IPV, such as Strength at Home, teach active listening and assertive messaging, and provide psychoeducation on common communication traps (Taft et al., 2015). Our findings align with the notion that psychoeducation on communication and practice of these skills might spur more productive communication during contentious conversations. Prior to delivering interventions, it may be useful to use either a self-report measure of disengagement (Barry et al., 2008) or conduct observations while taking note of indicators of disengagement such as couple eye contact behaviors, question asking, and listening behavior. In treatment, couples may benefit from being instructed to bring mindful awareness to their question asking, eye contact, and listening behaviors when discussing areas of conflict. Still, it is worth mentioning that even in situations in which lower levels of IPV is present, providing victims with conversational strategies to avoid future violence—as opposed to holding those who use violence against their partners responsible for their own behavior—may place undue blame on victims and is insufficient for addressing the problem (Holtzworth-Munroe & Meehan, 2004).
Future Directions
Future research should explore potential gender differences in the association between disengagement behaviors and IPV to better understand why women’s (but not men’s) disengagement behaviors might serve as a risk indicator of bidirectional IPV. Future research should also explore mechanisms that may explain why certain disengagement behaviors are related to IPV, keeping in mind that disengaged communication patterns likely emerged long ago in the context of unhealthy relationship dynamics. Another important area of inquiry is to investigate whether disengagement patterns differentially predict IPV across couples differing in important identity characteristics such as racial/ethnic background, nationality, and sexual and gender diversity. Indeed, norms around communication and patterns of engagement are markedly different across cultures and remain an important future direction for research on couple communication and IPV. Finally, one of the strengths of this study was the use of observational methods to examine disengagement behaviors, which participants may be less adept at recognizing and reporting on self-report measures. To extend this work, future research could measure these behaviors in couples’ naturalistic home environments via ecological momentary assessment techniques to increase external validity and better understand associations between disengagement behaviors and IPV as they unfold in real time.
Supplemental Material
sj-docx-1-jiv-10.1177_08862605231163644 – Supplemental material for Women’s Disengagement Behaviors During Couple Conflict: Investigating Risk Indicators of Intimate Partner Violence
Supplemental material, sj-docx-1-jiv-10.1177_08862605231163644 for Women’s Disengagement Behaviors During Couple Conflict: Investigating Risk Indicators of Intimate Partner Violence by Molly R. Franz, Madeline Smethurst, Robin A. Barry, Hannah E. Cole and Casey T. Taft in Journal of Interpersonal Violence
Supplemental Material
sj-docx-2-jiv-10.1177_08862605231163644 – Supplemental material for Women’s Disengagement Behaviors During Couple Conflict: Investigating Risk Indicators of Intimate Partner Violence
Supplemental material, sj-docx-2-jiv-10.1177_08862605231163644 for Women’s Disengagement Behaviors During Couple Conflict: Investigating Risk Indicators of Intimate Partner Violence by Molly R. Franz, Madeline Smethurst, Robin A. Barry, Hannah E. Cole and Casey T. Taft in Journal of Interpersonal Violence
Footnotes
Author Note
The views expressed in this article are those of the authors and do not reflect the official policy or position of the Department of the Army or Department of Veterans Affairs.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
Funding for this study was provided by the National Institute of Mental Health Award K08 MH73117 01A2 to Casey T. Taft.
Supplemental Material
Supplemental material for this article is available online.
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
