The dyadic construction of romantic conflict recovery sabotage

Abstract

This longitudinal study of 100 couples assessed individual and dyadic processes associated with romantic conflict recovery or how couples behave in the moments following conflict. Couples completed measures of attachment anxiety and avoidance; a conflict discussion during which affect, behavior, and conflict resolution were coded; a cooldown discussion during which post-conflict behavior was coded; and measures of relationship satisfaction and stability one year later. Recovery sabotage (negative behavior and perseveration on conflict in the moments following conflict) was associated with high attachment anxiety and low avoidance. Recovery sabotage was unrelated to affect expressed during conflict and was instead tied to whether partners aired or suppressed grievances. Consistent with the demand–withdraw conflict pattern, recovery sabotage was associated with lower actor conflict avoidance but higher partner conflict avoidance. These effects were independent of conflict resolution, which was not significantly associated with recovery sabotage when other features of conflict were controlled. Recovery sabotage and conflict resolution also differentially predicted satisfaction and stability one year later. Findings suggest recovery sabotage is a distinct, developmentally organized relationship process tied to attachment history and behavioral, rather than affective, transactions between partners during conflict.

Keywords

Attachment conflict avoidance romantic conflict recovery satisfaction

Romantic conflict begins when partners discover a discrepancy in their goals, needs, or motives that exceeds their set comfort level (Berscheid & Ammazzalorso, 2001). To maintain the relationship, partners must discuss the discrepancy (i.e., engage in conflict) until they eliminate or reduce it to a mutually tolerable level (i.e., achieve conflict resolution) and then let go of conflict to resume typical interaction patterns (i.e., conflict recovery). The first steps in this sequence, conflict and conflict resolution, have long been focal points in marital research. Both have well-documented origins in developmental history (Simpson, Collins, Tran, & Haydon, 2007; Whitton et al., 2008) and well-documented effects on relationship satisfaction and stability (Gottman & Levenson, 1992; Hojjat, 2000). However, the romantic conflict literature has tended to overlook conflict recovery or how couples transition out of conflict and resume other non-conflict interactions. Whether partners regain their psychological footing after conflict may make the difference between a righted ship, a downward spiral, or preservation of a miserable status quo. Yet despite its potential importance, conflict recovery is understudied (Ha, Overbeek, Lichtwarck-Aschoff, & Engels, 2013; Prager, Shirvani, Poucher, Cavallin, Truong, & Garcia, 2015) and little is known about antecedents, concurrent correlates, and outcomes associated with effective versus poor conflict recovery.

The emerging conflict recovery literature

The few existing studies of conflict recovery vary widely in their methods, target populations, and conceptualizations of conflict recovery. In perhaps the earliest study of how couples rebound after a conflict, Gottman and Levenson (1999) posited that couples who prevented negative affect generated during conflict from carrying forward to a subsequent conversation were more likely to remain together. Indeed, Gottman and Levenson (1999) successfully predicted divorce from negative affect expressed during a post-conflict “rebound” discussion. More recently, Ha et al. (2013) conceptualized conflict recovery as positive emotion expressed following conflict, but post-conflict positive affect did not predict breakup in their sample of adolescent couples.

Taking a different approach, Salvatore and colleagues (Salvatore, Kuo, Steele, Simpson, & Collins, 2011) conceptualized conflict recovery as a behavioral self-regulatory process informed by developmental history. Using data from a longitudinal study of over 20 years, they observed post-conflict behavior when young adult target participants (followed since birth) and their romantic partners were asked to transition from conflict to a discussion of agreements. They defined effective recovery as nominating areas of agreement and elaborating the partner’s positive contributions, while poor recovery was defined as continuing to discuss the prior conflict, bringing up new disagreements, or negating the partner’s positive contributions about areas of agreement. Target participants’ post-conflict behavior in early adulthood had roots in infant attachment history measured 20 years earlier. Infant attachment security was associated with effective recovery, while infant insecurity was associated with poor recovery. Furthermore, the relationships of people with insecure attachment histories were buffered (i.e., more stable) when their partners recovered well from conflict.

Most recently, in contrast to observational studies of conflict recovery, Prager and colleagues (Prager et al., 2015) studied the psychological experience of conflict recovery using daily diary assessments. They assessed self-reported conflict, positive and negative affect, satisfaction, and intimacy over three weeks; post-conflict rebound of satisfaction, affect, and intimacy served as indicators of conflict recovery. Consistent with Salvatore, et al.’s (2011) findings, attachment insecurity was associated with reduced same-day and day-after-conflict positive affect; partners of avoidant and anxious people also reported feeling worse following days on which conflict occurred than did partners of secure people.

These studies suggest that conflict recovery has roots in developmental history and that the extent to which partners recover from conflict impacts the course of their relationships. Yet, several aspects of conflict recovery warrant additional research. Specifically, more research is needed to build on evidence of links between conflict recovery and attachment, to test for the first time whether conflict recovery is associated with what happened in the preceding conflict, to take a dyadic perspective on antecedents and outcomes of conflict recovery, and to clarify mixed findings about links between conflict recovery and important outcomes. Below, we elaborate the rationale for these priorities.

Research priorities

Links to conflict processes

Very little is known about whether conflict recovery is associated with what happened during the preceding conflict. We conceptualize at least three interrelated explanations for how conflict processes may relate to subsequent conflict recovery. The first is that poor recovery reflects affect spillover from the preceding conflict: negative affect generated during conflict lingers in the moments after conflict, precluding effective recovery (Gottman & Levenson, 1999). However, no prior study has tested whether individual affect expressed during conflict is associated with conflict recovery behavior. Second, conflict recovery is likely contingent on both partners’ behavior during conflict. For example, maladaptive strategies employed during conflict (e.g., stonewalling, withdrawal, criticism, or complaint) may render effective recovery more difficult for one or both partners. However, no prior study has examined conflict recovery vis-à-vis the specific behavioral strategies used to engage in conflict. Third, post-conflict behavior may depend on conflict resolution. That is, effective recovery may simply be the afterglow of a resolved conflict, while those with lingering unresolved discrepancies cannot recover well and may experience rumination (Bevan, Hefner, & Love, 2014). These three possibilities are not mutually exclusive and all may contribute to post-conflict behavior. However, none has been fully tested with appropriate controls in place and there is little extant theory about how effective or poor recovery might be associated with what transpired between partners during conflict or whether couples reached a resolution. Thus, whether recovery is distinct from or related to these aspects of conflict is unknown.

Links to attachment

Attachment history appears to be a key source of variation in post-conflict behavior (Prager et al., 2015; Salvatore et al., 2011). Attachment theory suggests that highly anxious individuals show poor conflict recovery due to the relatively protracted distress and rumination they tend to experience in response to perceived negative relationship events (Mikulincer, Shaver, & Pereg, 2003). By contrast, avoidant individuals, who tend to defensively minimize negative experiences in close relationships (Mikulincer et al., 2003), may not show poor recovery and may even appear to recover well. To date, only Prager et al. (2015) have examined conflict recovery with respect to adult attachment anxiety and avoidance. More research is needed to build on these findings. In addition, because attachment, emotion expression, conflict behavior, and conflict resolution are closely related processes (Simpson et al., 2007), it is important to test attachment alongside these other potential sources of variation to determine whether they exert distinct or overlapping influence on conflict recovery behavior.

Dyadic perspective

Existing evidence points to the importance of taking a dyadic perspective on conflict recovery. Having an insecure partner slowed affective recovery from conflict (Prager et al., 2015), while effective partner recovery buffered relationships of people with insecure attachment histories (Salvatore et al., 2011). Beyond these results, however, a dyadic perspective on behavioral and affective antecedents of conflict recovery—in which both actor effects (the effect of one’s own characteristics on one’s own outcome) and partner effects (the effects of the partner’s characteristics on one’s own outcome) on recovery are modeled—is lacking. Having an insecure partner or a partner who employs unconstructive behavior or expresses high levels of negative affect during conflict may make it more difficult for both partners to recover, yet no prior study has examined the influence of one partner’s conflict affect or behavior on the other’s recovery. To better understand individual recovery in the dyadic context in which it occurs, it is necessary to identify both actor and partner effects on post-conflict behavior.

Predicting relationship satisfaction and stability

Finally, information about whether and how conflict recovery predicts relationship outcomes such as satisfaction and relationship stability is mixed. Past research has produced conflicting findings about whether conflict recovery predicts stability (Gottman & Levenson, 1999; Ha et al., 2013; Salvatore et al., 2011). Additionally, none of these studies accounted for whether couples effectively resolved conflict, leaving open the possibility that links to stability were attributable to conflict resolution. Furthermore, no prior study of conflict recovery has examined prospective links to satisfaction. Relationship dissolution is thought to be independent of declines in satisfaction to some extent (Karney & Bradbury, 1995), but more research is needed to identify the distinctive contributions of different relationship processes to each outcome (Gottman, 2014). For example, it is possible that conflict resolution and recovery are distinct processes that independently affect satisfaction and stability. Testing this possibility is an important next step for establishing whether conflict recovery behavior is a distinct relationship process with long-term consequences.

The current study

We addressed these research priorities in a longitudinal study of a diverse community sample of 100 couples. Prior research suggests that conflict recovery is a multifaceted process that involves containment of negative affect and behavior as well as the presence of positive affect and behavior. However, because decades of relationship research indicate that “not being nasty matters more than being nice” (Ewart, Taylor, Kraemer, & Agras, 1991), we were particularly interested in understanding correlates and consequences of negative recovery behavior. Specifically, building on Salvatore et al.’s (2011) conceptualization of conflict recovery, we assessed recovery sabotage, or the extent to which people actively harped on the prior conflict, dredged up new disagreements, or negated their partner’s positive contributions during a post-conflict rebound conversation. These negative behaviors stand out as particularly aberrant in the context of the laboratory task where couples are directed to discuss areas of agreement, and the emergence of recovery sabotage under these circumstances seems particularly pernicious and meaningful in the context of a post-conflict opportunity for rebound and repair.

With respect to the research priorities articulated above, we sought to build on existing links between conflict recovery and attachment. We opted to assess adult attachment avoidance and anxiety with respect to parents to build on the link between infant attachment to caregivers and recovery behavior observed by Salvatore et al. (2011) and test for the first time whether recovery sabotage is associated with adults’ representations of their current relationships with caregivers. Based on prior research and theory (Prager et al., 2015; Salvatore et al., 2011), we expected higher recovery sabotage would be associated with low attachment avoidance and high anxiety (Hypothesis 1). Based on Prager et al.’s (2015) finding that people with insecure partners were negatively affected by conflict, we also expected that higher actor recovery sabotage would be associated with higher partner avoidance and anxiety (Hypothesis 2).

We also studied attachment together with facets of conflict. In line with the affect spillover hypothesis (Gottman & Levenson, 1999), we tested whether actor and partner positive and negative affect expressed during conflict were associated with recovery sabotage. We expected higher recovery sabotage to be associated with higher actor negative affect and lower actor positive affect during conflict (Hypothesis 3). In light of emotion contagion theory (Hatfield, Cacioppo, & Rapson, 1994), we also expected that higher partner negative affect and lower partner positive affect would be associated with actor recovery sabotage (Hypothesis 4).

To test whether behavioral strategies employed during conflict are associated with post-conflict behavior, we examined two especially damaging behavioral strategies. The first, referred to here as negative engagement, involves attack, blame, and being especially critical or harsh (Gottman, 1993). We expected that individuals who negatively engage during conflict or those with a partner who negatively engages are likely to sabotage recovery because negative engagement may exacerbate the perceived severity of a conflict and make recovery more difficult (Hypothesis 5). The second behavioral strategy, conflict avoidance, involves efforts to avoid disagreements by minimizing problems or withdrawing verbally and behaviorally during the interaction (Gottman, 1993). We expected that actor versus partner conflict avoidance would have divergent effects on recovery sabotage. People who avoid conflict may not sabotage recovery because the post-conflict period affords an opportunity consistent with their implicit goal to sidestep conflict. Conversely, drawing on what is known about the demand–withdraw pattern (Caughlin & Vangelisti, 2000), having a partner who goes out of the way to avoid conflict by withdrawing may trigger a “demand” response in the other partner that makes it harder to recover from conflict. Thus, we expected recovery sabotage would be negatively associated with actor conflict avoidance but positively associated with partner conflict avoidance (Hypothesis 6).

With respect to whether conflict recovery is a distinct relationship process, we expected that post-conflict behavior would be independent of conflict resolution because, theoretically, it would be adaptive for couples to recover even if conflict was not yet resolved. For most couples trying to balance work, family, social, health, and other obligations, engaging in conflict without interruption until reaching resolution is not always feasible (Bevan, 2014). More often, conflicts arise when there is insufficient time to resolve them because one or both partners need to attend to other competing demands (e.g., childcare, work responsibilities). In such cases, we argue, it would be adaptive to be able to “pause” a conflict in progress, even if unresolved, and recover in the short-term to turn to other pressing relationship or individual needs. Thus, how partners behave after conflict should be independent of whether or not conflict was resolved. We expected that conflict resolution would not be significantly associated with recovery sabotage when other features of conflict (affect and behavior) were controlled (Hypothesis 7), and that the effects of attachment, affect, and behavior proposed above would hold when conflict resolution was statistically controlled (Hypothesis 8).

Finally, we examined whether recovery sabotage predicted two key longitudinal outcomes: satisfaction and stability. We suspected that recovery sabotage and conflict resolution would play distinct roles in satisfaction versus stability outcomes. Overtime, unresolved issues may ultimately erode relationships no matter how well partners recover in between conflict episodes (Bevan, Finan, & Kaminski, 2008). We therefore expected that conflict resolution would be a stronger predictor of stability than recovery sabotage (Hypothesis 9). Conversely, couples in which one or both partners perseverate on conflict are likely to be less satisfied because perseveration interferes to some extent with other types of interactions that promote relationship satisfaction (e.g., intimacy, caregiving, and shared interests), whereas effective recovery should allow partners to contain conflict and resume other parts of their relationship, thereby promoting satisfaction while they continue to work out unresolved conflicts. This prediction is supported by evidence that unresolved conflicts do not necessarily undermine relationship satisfaction and can lead to positive motivation to make additional attempts at conflict resolution (Bevan et al., 2014). Therefore, we expected recovery sabotage would be a stronger predictor of long-term satisfaction than conflict resolution (Hypothesis 10).

Method

Participants

One hundred couples were recruited from communities in western New England by paper and electronic posts to community bulletin boards. Sample size was determined based on effect sizes reported in prior research, which typically range from r = .20 to .30. Power analysis, adjusted for non-independence between partners according to recommendations in Kenny, Kashy, and Cook (2006), indicated that a sample size between 69 and 98 couples would be sufficient. Data collection stopped when the sample reached 100 couples. Participants were 19–43 years old (mean = 26.8 years, SD = 5.3); couples had been involved for at least 1 year (mean relationship length = 3.2 years, SD = 2.6 years). Participants were fairly diverse with respect to ethnicity, gender, sexual orientation, and socioeconomic status. When asked to report ethnic background(s), 79.5% identified as having European heritage, 13% as Latina/Latino, 7% as Native American, 5% as Asian, 2.5% as Middle Eastern/Arab, and 2.5% as Black or African; 15% identified multiple ethnic heritages and 5.5% did not report an ethnic background. Forty-two percent had a bachelor’s degree or higher; 52.5% reported an annual household income of less than USD$50,000 per year. The sample included 90 men, 107 women, 3 participants who identified as transgender or genderqueer, 11 lesbian couples, 3 gay couples, and 2 couples in which one or both partners identified as transgender or genderqueer. Fifty percent of couples were married or committed, 10% were engaged, and 40% were dating exclusively.

Procedure

Before the laboratory visit, participants completed online attachment and relationship perception measures. Couples then completed a 2-hr laboratory session. Research assistants facilitated application of sensors to record heart rate, skin conductance, and respiration; physiological data are not discussed in this report. Participants then completed the Markman–Cox Conflict Discussion Task (Cox, 1991) in which they discussed their biggest disagreement for 10 min. Immediately following this task, couples completed a “cool-down” task in which they were directed to discuss areas of greatest agreement for 5 min; this procedure is identical to that used by Salvatore et al. (2011). Participants then completed a joint interview about their relationship’s history (also not discussed in this report). Participants were debriefed and compensated USD$20 each. The complete procedure was approved by the Mount Holyoke College Institutional Review Board.

Measures

Relationship satisfaction

Time 1 relationship satisfaction was assessed during the initial online survey via the 7-item Relationship Assessment Scale (RAS; Hendrick, 1988). Cronbach’s α was .80.

Attachment avoidance and anxiety

The experiences in close relationships–relationship structures measure (ECR-RS; Fraley, Heffernan, Vicary, & Brumbaugh, 2011) assessed avoidance and anxiety for mother- and father-like attachment figures. We assessed adult attachment to caregivers to conceptually parallel infant attachment to caregivers, which was associated with conflict recovery behavior in the Salvatore et al.’s (2011) study. Participants rated the same 10 items on a 7-point scale for mother and father. Avoidance and anxiety dimensions were computed for each figure; scales were then averaged into composite avoidance and anxiety scales according to the procedure described by Fraley, Heffernan, Vicary, and Brumbaugh (2011). Cronbach’s α were .91 for mother anxiety, mother avoidance, and father anxiety and .90 for father avoidance. Mother and father avoidance scales were correlated at r = .25, p < .001; mother and father anxiety scales were correlated at r = .35, p < .001.

Positive and negative affect

Two trained raters scored participants on two 5-point scales that assessed positive and negative affect during the 10-min conflict discussion. These individual-level scales were adapted from dyad-level scales used to code shared positive and negative affect (Collins et al., 1999). High scores on positive affect were assigned to participants who displayed genuine smiling, laughing, or physical or verbal affection, while low scores were assigned to those who did not display these signs of positive affect. High scores on negative affect were assigned to participants who displayed overt signs of annoyance, frustration, tension, anger, or hostility; low scores were assigned to those in which such expressions were absent. Interrater reliability (one-way random intraclass correlations [ICC]) established on 23% of the sample was high for both scales: positive affect = .92 and negative affect = .90.

Conflict behavior

Two trained raters scored participants on two 5-point scales that assessed conflict behavior. Conflict avoidance assessed the extent to which participants deflected, avoided, or withdrew from conflict by skirting the topic, attempting to change the subject, or otherwise appearing reluctant to engage in conflict. Scores of 5 were assigned to those who actively avoided conflict throughout the discussion without employing other behavioral strategies; participants who avoided conflict through most of the interaction but displayed one or two instances of other strategies were rated 4; participants who avoided conflict at several points during the interaction but directly addressed conflict at other times were scored 3; participants who displayed only one or two minor instances of avoidance were scored 2; and those who did not avoid conflict at any point were scored 1. Inter-rater reliability (ICC) was .88.

Negative engagement assessed the extent to which participants criticized or blamed their partners, complained, or made negative demands for change in their partners. Participants who actively used blame, criticism, or negative demands consistently throughout the conflict and did not employ other strategies were scored 5; participants who predominantly used negative engagement but showed one or two instances of other behavioral strategies were scored 4; participants who balanced instances of criticism, blame, or demand with other behavioral strategies were scored 3; participants who displayed only one or two minor, fleeting instances of negative engagement were scored 2; and participants who did not negatively engage at any time were scored 1. Inter-rater reliability (ICC) was .96.

Conflict resolution

The same raters coded each couple’s dyadic conflict resolution, which assessed on a 7-point scale the extent to which the couple effectively resolved conflict (Collins et al., 1999). High scores were assigned to couples in which partners effectively collaborated to achieve a mutually satisfying resolution; moderate scores were assigned to couples for whom the process was somewhat strained or tense at times and led to a less than satisfactory outcome; low scores were assigned to couples in which only one partner or neither partner appeared satisfied with the outcome or couples in which partners did not reach a solution of any kind. Inter-rater reliability (ICC) was .90.

Recovery sabotage

Two additional trained coders who did not rate conflict affect, behavior, or resolution rated each participant’s behavior during the cool-down discussion on a 5-point Likert-type recovery sabotage scale. We developed this scale to capture the lower end of Salvatore et al.’s (2011) original conflict recovery scale. Scores of 5 were assigned when individuals consistently sabotaged the post-conflict interaction by perseverating on the preceding conflict, bringing up new problems, disputing their partner’s suggestions of topics on which the couple agreed, or refusing to talk; participants for whom sabotage was a predominant theme of the interaction but who made a few fleeting attempts to discuss agreements were scored 4; participants who sabotaged the interaction at times, but at other times did not make negative contributions were scored 3; participants who made only a few minor attempts at sabotage were scored 2; and participants who did not make any negative contributions were scored 1. This does not necessarily mean that the individual contributed positively to the discussion; however, the individual did not bring up topics on which the couple disagreed or negate the positive contributions (if any) of the partner. Inter-rater reliability (ICC) was .93.

Time 2 satisfaction and stability

Couples were contacted by email 1 year after their lab visit to assess relationship stability and (if the couple remained together) each partner’s satisfaction. Partners were contacted separately to ensure confidentiality and to obtain independent responses from partners regardless of relationship status. Stability was assessed by asking whether couples remained together or had broken up over the past year. Of the 100 couples, we obtained information on 70; of these, 53 remained together and 17 had ended their relationships. Participants who responded and were still involved with their Time 1 partners (n = 67) also completed a brief 3-item version of the RAS (Hendrick, 1988) comprising the 3 items with the strongest loadings when all 7 items were submitted to factor analysis: “How well does your partner meet your needs?” “How good is your relationship compared to most?” and “In general, how satisfied are you with your relationship?” Items were highly intercorrelated (r ranged from .49 to .77; p < .001). The mean of these items served as the Time 2 satisfaction score; Cronbach’s α for the RAS short version was .81.

We observed fairly high attrition rates for the Time 2 assessment of stability (30 couples; 30% of the original sample) and satisfaction (73 individuals; 52% of individuals in known intact couples at Time 2). Women (62%) were more likely than men (24%) to respond at Time 2, χ²(1) = 24.15, p < .001. Of the 67 participants in known intact relationships at Time 2 who reported satisfaction data, 48% were people whose partner also responded at Time 2. Time 2 response rate was higher among same-sex or queer-identified couples (81%) than for heterosexual couples (68%), but this difference was not statistically significant, χ²(1) = 1.15, p = .28. The difference in Time 2 response rate for married (74%) versus unmarried couples (67%) was also not statistically significant, χ²(1) = 0.62, p = .43. Independent samples t-tests showed that Time 2 non-responders did not differ significantly from those who completed the Time 2 assessment on any Time 1 variables; thus, it appears that the subsample that completed the Time 2 assessment was largely representative of the original sample.

Results

Analytic plan

According to best practices for dyadic data analysis, we first tested whether heterosexual dyads (84 of 100 couples) were distinguishable by sex (Kenny, Kashy, & Cook, 2006). An omnibus test of distinguishability computed with Kenny’s (2015) online Dingy application indicated that heterosexual dyads were not distinguishable by sex, χ²(42) = 46.65, p = .29; all analyses treated dyads as indistinguishable. To test for non-independence of indistinguishable partners’ data, we computed partial ICC between partners on all variables according to the method described in Kenny et al. (2006); ICCs for all behavioral measures were significant at p < .001, indicating non-independence. We tested Hypotheses 1–9 using actor–partner interdependence models (APIMs) computed in SPSS version 22; no random effects were specified. APIM assesses the effect of each person’s own scores on his or her outcome (actor effects) as well as the effects of the partner’s scores on his or her outcome (partner effects), controlling for the non-independence between partners. Where relevant, models controlled for relationship length because of its associations with other variables in the model. All variables were standardized prior to model testing. B estimates serve as indicators of effect size; 95% confidence intervals are included in tables. Means, standard deviations, and direct associations among all variables appear in Table 1.

Table 1.

Descriptive statistics and unstandardized coefficients from APIMs testing direct associations among study variables.

	1	2	3	4	5	6	7	8	9	10	Mean	SD
1. Recovery sabotage	–										2.16	1.12
2. Attachment anxiety	.14^†	–									2.00	1.25
3. Attachment avoidance	−.11	.80***	–								3.56	1.30
4. Positive affect	−.07	−.03	−.06	–							3.05	1.13
5. Negative affect	.20*	.11	−.05	−.23**	–						1.46	0.84
6. Conflict avoidance	−.18^*	−.01	.08	−.14^*	.10	–					1.41	0.86
7. Negative engagement	.23**	.27**	−.02	−.25***	.49***	.12*	–				1.39	0.91
8. Conflict resolution	−.28**	−.17^†	.00	.26***	−.42***	−.45***	−.49***	–			4.71	1.54
9. Time 1 satisfaction	−.07	−.01	−.11	.23^**	−.10^†	−.06	−.15*	.14^†	–		5.71	0.68
10. Time 2 satisfaction	−.18	−.23	−.26	.28	−.10	.03	−.08	−.08	.22^†	–	6.31	0.57
11. Relationship length	.09	.02	.03	−.08	−.05	−.10	−.15^†	.22**	−.12*	−.17*	3.24	2.62
Partner recovery sabotage	.44***	−.14	.03	−.20**	−.15*	.15**	.06
Partner attachment anxiety	−.05	.10	.12^†	−.18*	−.02	.08	−.02
Partner attachment avoidance	−.02	.05	.14	−.12	−.07	.07	−.06
Partner positive affect	−.21^†	−.19	−.10	.80***	−.05	.05	−.12
Partner negative affect	.09	−.08	−.10	−.16*	.51***	.13*	.11*
Partner conflict avoidance	.27**	.12	.07	.01	.12^†	.65***	.31***
Partner negative engagement	.06	−.14	−.09	−.22**	.12*	.27***	.46***

Note. N = 100 couples (200 individuals). Within-person associations appear in the top portion of the table; actor–partner associations appear in the lower portion; coefficients given in italics are intraclass correlations that express associations between indistinguishable partners’ scores on the same measure (Kenny et al., 2006). APIMs = actor–partner interdependence models.

^† p < .10; *p < .05; **p < .01; ***p < .001.

Actor and partner attachment anxiety and avoidance

To test our prediction that recovery sabotage would be associated with adult representations of parental attachment, we examined whether recovery sabotage was associated with actor (Hypothesis 1) and partner (Hypothesis 2) parental attachment avoidance and anxiety (Table 2). Results revealed that individuals who reported more attachment anxiety (b = .32, p = .001) and less attachment avoidance (b = −.30, p = .002) showed higher recovery sabotage; Hypothesis 1 was supported. Partner attachment avoidance and anxiety were not associated with actor recovery sabotage; Hypothesis 2 was not supported.

Table 2.

Actor and partner attachment anxiety and avoidance predicting recovery sabotage.

						CI₉₅
Fixed effects	Estimate	SE	df	t	p	Lower	Upper
Intercept	−.01	.09	94	−.07	.95	−.18	.16
Actor attachment anxiety	.32***	.09	146	3.39	.00	.13	.50
Actor attachment avoidance	−.30***	.09	154	−3.21	.00	−.49	−.11
Partner attachment anxiety	.01	.09	145	.08	.94	−.18	.19
Partner attachment voidance	−.04	.09	154	−.38	.71	−.22	.15
Relationship length	.09	.09	97	1.00	.32	−.08	.25

Note. N = 100 couples (200 individuals). SE = standard error.

***p < .001.

Actor and partner affect during conflict

To test the affect spillover prediction that recovery sabotage would be associated with higher negative and lower positive affect expressed during conflict by actors (Hypothesis 3) and partners (Hypothesis 4), we tested a model that included actor and partner positive and negative affect as predictors of recovery sabotage. As shown in Table 3, recovery sabotage was not significantly associated with actor positive affect (b = −.03, p = .74), actor negative affect (b = .14, p = .07), partner positive affect (b = −.17, p = .09), or partner negative affect (b = .05, p = .54). Hypotheses 3 and 4 were not supported.

Table 3.

Actor and partner conflict affect predicting recovery sabotage.

						CI₉₅
Fixed effects	Estimate	SE	df	t	p	Lower	Upper
Intercept	−.01	.08	94	−.10	.93	−.17	.16
Actor positive affect	−.03	.10	151	−.33	.74	−.23	.16
Partner positive affect	−.17^†	.10	151	−1.71	.09	−.36	.03
Actor negative affect	.14^†	.08	190	1.85	.07	−.01	.29
Partner negative affect	.05	.08	190	.62	.54	−.10	.20
Relationship length	.07	.09	97	.88	.38	−.09	.24

Note. N = 100 couples (200 individuals). SE = standard error.

^† p < .10.

Actor and partner behavior during conflict

Next, we tested the prediction that higher recovery sabotage would be associated with higher actor and partner negative engagement during conflict (Hypothesis 5) and lower actor but higher partner conflict avoidance (Hypothesis 6). As shown in Table 4, actors who showed more negative engagement displayed higher recovery sabotage (b = .17, p = .03) but partner negative engagement was not significantly associated with recovery sabotage (b = .14, p = .07); Hypothesis 5 was partially supported. In addition, we observed the expected divergence in the effects of actor versus partner conflict avoidance on recovery sabotage. Actors displayed more recovery sabotage when their own behavior was less conflict avoidant (b = −.24, p = .01) and when their partner’s behavior was more conflict avoidant (b = .21, p = .02); Hypothesis 6 was supported. To probe the robustness of these behavioral effects when affect and behavior were tested simultaneously, we tested an additional model that included all actor and partner affect and conflict behavior variables. The effects of actor conflict avoidance (b = −.24, p = .01) and partner conflict avoidance (b = .21, p = .02) did not differ from those observed previously and were the only significant predictors of recovery sabotage; actor negative engagement was no longer significantly associated with recovery sabotage. Hypothesis 6 was supported under these more rigorous conditions, but Hypothesis 5 was not.

Table 4.

Actor and partner conflict behavior predicting recovery sabotage.

						CI₉₅
Fixed effects	Estimate	SE	df	t	p	Lower	Upper
Intercept	−.01	.08	94	−0.08	.95	−.17	.16
Actor conflict avoidance	−.24*	.09	183	−2.83	.01	−.41	−.07
Partner conflict avoidance	.21*	.09	183	2.41	.02	.03	.37
Actor negative engagement	.17*	.08	189	2.20	.03	.02	.32
Partner negative engagement	.14^†	.08	189	1.83	.07	−.01	.29
Relationship length	.13	.09	98	1.48	.14	−.04	.30

Note. N = 100 couples (200 individuals). SE = standard error.

*p < .05; ^† p < .10.

Recovery sabotage and conflict resolution

Next, we tested our prediction that conflict resolution would not be significantly associated with recovery sabotage when other facets of conflict were controlled (Hypothesis 7) and that effects observed for attachment and conflict behavior on recovery sabotage would be independent of conflict resolution (Hypothesis 8). The model included actor attachment avoidance and anxiety, actor and partner behavioral conflict avoidance and negative engagement, and dyadic conflict resolution (Table 5). We did not include partner attachment anxiety and avoidance or actor and partner positive and negative affect because of the null results observed in previous models.¹ Dyadic conflict resolution was not significantly associated with recovery sabotage (b = −.22, p = .07); Hypothesis 7 was supported.

Table 5.

Attachment, affect, behavior, and conflict resolution predicting recovery sabotage.

						CI₉₅
Fixed effects	Estimate	SE	df	t	p	Lower	Upper
Intercept	−.01	.08	93	−0.07	.95	−.17	.16
Attachment anxiety	.27***	.08	143	3.42	.00	.11	.42
Attachment avoidance	−.24***	.08	148	−3.06	.00	−.40	−.09
Actor conflict avoidance	−.27***	.09	186	−3.13	.00	−.44	−.10
Partner conflict avoidance	.18*	.09	186	2.07	.04	.01	.35
Actor negative engagement	.04	.08	178	0.48	.63	−.12	.20
Partner negative engagement	.09	.08	177	1.06	.29	−.07	.25
Conflict resolution	−.22^†	.12	94	−1.86	.07	−.45	.02
Relationship length	.14	.08	97	1.66	.10	−.03	.31

Note. N = 100 couples (200 individuals).SE = standard error.

^† p < .10; *p < .05; ***p < .001.

In addition, when conflict resolution was controlled, individuals displayed higher recovery sabotage when they were higher in attachment anxiety (b = .27, p = .001) or lower in attachment avoidance (b = −.24, p = .003), when they engaged in less conflict avoidance behavior (b = −.27, p = .002) or when their partners engaged in more conflict avoidance behavior (b = .18, p = .04). Actor negative engagement behavior (b = .04, p = .63), partner negative engagement behavior (b = .09, p = .29), and relationship length (b = .14, p = .10) were not significantly associated with recovery sabotage. Hypothesis 8 was supported.

Satisfaction and stability outcomes

Finally, we tested whether recovery sabotage and conflict resolution differentially predicted satisfaction and stability overtime. For each outcome, we tested the effect of conflict resolution on the outcome in Step 1, followed by a second step that included recovery sabotage. This strategy enabled us to test whether conflict resolution and recovery sabotage differentially predicted each outcome when the other was controlled and rule out the possibility that conflict resolution indirectly affected each outcome through recovery sabotage. First, we tested our prediction that actor and partner recovery sabotage at Time 1 would be more strongly associated with lower satisfaction at Time 2 than was Time 1 conflict resolution (Hypothesis 9), controlling for Time 1 actor satisfaction (Table 6). In Step 1, neither conflict resolution (b = −.02, p = .90) nor Time 1 satisfaction (b = .15, p = .29) were significantly associated with Time 2 satisfaction. In Step 2, actors whose partners displayed more recovery sabotage at Time 1 were less satisfied 1 year later (b = −.27, p = .02). Time 2 satisfaction was not significantly associated with actor recovery sabotage (b = .09, p = .46), dyadic conflict resolution (b = −.04, p = .81), or actor Time 1 satisfaction (b = .08, p = .57). Hypothesis 9 was partially supported in that partner but not actor recovery sabotage predicted Time 2 satisfaction better than conflict resolution.

Table 6.

Actor and partner recovery sabotage predicting actor Time 2 satisfaction.

						CI ₉₅
Fixed effects	Estimate	SE	df	t	p	Lower	Upper
Step 1
Intercept	−.02	.15	43	−0.16	.87	−.33	.28
Conflict resolution	−.02	.16	41	−0.12	.90	−.33	.30
T1 satisfaction	.15	.14	43	1.06	.29	−.13	.43
Step 2
Intercept	−.02	.15	43	−0.12	.91	−.32	.28
Conflict resolution	−.04	.16	41	−0.24	.81	−.36	.28
T1 satisfaction	.08	.13	38	0.57	.57	−.19	.34
Actor recovery sabotage	.09	.12	56	0.74	.46	−.15	.33
Partner recovery sabotage	−.27*	.11	60	−2.46	.02	−.50	−.05

Note. N = 67 individuals. SE = standard error.

*p < .05.

To test our prediction that conflict resolution would predict 1-year stability better than recovery sabotage (Hypothesis 10), we computed a binary logistic regression using the procedure recommended for predicting dyad-level outcomes (e.g., stability) from individual-level predictors (e.g., each partner’s recovery sabotage score) in indistinguishable dyads (see Goldberg & Garcia, 2015). Using stepwise logistic regression (Table 7), 1-year relationship stability (dissolution = 0; stability = 1) was regressed on dyadic conflict resolution (Step 1) and the sum of partners’ recovery sabotage scores (Step 2). Higher conflict resolution was significantly associated with relationship stability in Step 1 (B = 0.75, p = .02, odds ratio = 2.12, omnibus χ² = 6.18, p = .01). In Step 2 (step χ² = 0.77, p = .38, omnibus χ² = 6.95, p = .03), conflict resolution remained significantly associated with stability (B = 0.70, p = .03, odds ratio = 2.01) but summed recovery sabotage was not significantly associated with stability (B = −0.26, p = .38, odds ratio = 0.78).² For every one unit increase in conflict resolution, couples were twice as likely to remain together 1 year later. In supplemental analyses, we tested a model using only the subset of heterosexual dyads (n = 84) in which the effects of recovery sabotage on stability could be modeled separately for men and women. Results for conflict resolution were unchanged (B = 0.75, p = .048, odds ratio = 2.11) and neither men’s nor women’s recovery sabotage significantly predicted stability (men: B = −0.33, p = .34, odds ratio = 0.72; women: B = −0.03, p = .94, odds ratio = 0.97). Hypothesis 10 was supported.

Table 7.

Conflict resolution and summed recovery sabotage predicting relationship stability.

	Step 1				Step 2
	B	SE (b)	Wald χ²	Odds ratio	B	SE (b)	Wald χ²	Odds ratio
Constant	1.20	.30	15.93	–	1.21	.30	15.79	–
Conflict resolution	0.75	.32	5.51	2.12*	0.70	.32	4.68	2.01*
Summed recovery sabotage	−0.26	.29	0.78	0.78

Note. N = 70 couples. SE = standard error.

*p < .05.

Discussion

To address research priorities in the emerging literature on conflict recovery, we examined associations of attachment, affect, behavior, and conflict resolution with recovery sabotage observed in the moments after conflict. We expected that recovery sabotage, a measure of attempts to sabotage opportunities to transition away from conflict toward other shared goals, would be associated with attachment and both partners’ affect and behavior during conflict. We also predicted that recovery sabotage would be unrelated to whether or not conflicts were resolved, stemming from the idea that couples should be able to recover momentarily from conflict in order to pursue other goals even if conflicts are not resolved (Bevan, 2015). Finally, anticipating that recovery sabotage plays an important role in relationship satisfaction but that conflict resolution is ultimately more important than recovery for relationship stability, we expected that recovery sabotage would predict long-term satisfaction better than conflict resolution and that conflict resolution would predict relationship stability better than recovery sabotage. These predictions were generally confirmed. Together, results document that recovery sabotage is co-constructed by romantic partners via individual and dyadic processes and is distinctively tied to how much partners air versus suppress grievances during conflict rather than whether those grievances are resolved. We highlight and discuss the key findings below.

Attachment and recovery sabotage

Consistent with prior evidence of links between attachment and conflict recovery (Prager et al., 2015; Salvatore et al., 2011), recovery sabotage was strongly associated with lower actor parental attachment avoidance and higher parental attachment anxiety. For highly anxious people, effective recovery appears to be rather difficult, perhaps reflecting the regulatory challenges associated with trying to contain emotional and behavioral responses to a potential relationship threat such as conflict (Feeney, 2004). Avoidance, on the other hand, was associated with less recovery sabotage. At first glance, this seems to suggest that avoidance is a protective factor. However, we view our finding as consistent with the tendency of avoidant individuals to dismiss and disengage from relationship threats (Mikulincer et al., 2003), to suppress positive and negative affect in romantic interactions (Haydon, Roisman, & Burt, 2012), and to use behavioral withdrawal to cope with physiological stress reactivity to conflict (Gottman & Levenson, 1986). While the use of deactivating strategies appears to protect avoidant individuals from actively sabotaging the moments after conflict, prior evidence that infant and adult attachment insecurity are associated with low positive rebound from conflict (Prager et al., 2015; Salvatore et al., 2011) suggests that attachment avoidance also prevents people from affectively rebounding from conflict and contributing to opportunities for repair that benefit both partners.

Despite prior evidence that people whose partners had insecure romantic attachment fared worse in the aftermath of conflict (Prager et al., 2015), partner attachment to parents was not associated with actor recovery sabotage. Nonetheless, this study extends prior evidence of links between infant attachment to caregivers and later recovery behavior in adult romantic relationships (Salvatore et al., 2011) and between adult self-reported attachment to romantic partners and the psychological experience of recovery (Prager et al., 2015). Links between attachment and conflict recovery have now been observed in three distinct samples, in different assessment periods (infant vs. adult attachment), using different measurement strategies (the Strange Situation Procedure vs. self-reported attachment style), and different attachment figures (parents vs. the current romantic partner). Together, these findings build the case that attachment history is a robust context from which the capacity for conflict recovery emerges and persists.

Affect spillover and behavioral strategies

Prior studies have posited that failure to rebound from conflict reflects affect spillover from conflict (Gottman & Levenson, 1999; Ha et al., 2013). In addition to testing the affect spillover hypothesis, we made the novel prediction that recovery sabotage would be closely tied to actor and partner behavior during conflict. Results indicated that each partner’s behavior during conflict, rather than the affective tone of conflict, predicted recovery sabotage. We found no evidence to support the hypothesis that high levels of negative affect during conflict render recovery more difficult or that high levels of positive affect during conflict preclude recovery sabotage. Recovery sabotage was also not significantly associated with either partner’s negative conflict behavior, such as criticism or complaint, when conflict resolution was controlled. By contrast, recovery sabotage was systematically associated with both partners’ conflict avoidance. These effects were independent of whether couples resolved conflict, suggesting that the strategies partners use to engage in conflict, rather than the affective tone of conflict or whether conflict is resolved, paves the way for recovery sabotage.

That conflict avoidance rather than negative engagement was associated with recovery sabotage may reveal even more about why sabotage occurs. Recovery sabotage seems to reflect difficulty transitioning out of conflict, not due to uncontained negative affect that arises from engaging in conflict, but rather uncontained negative affect that arises from failure to engage in conflict (i.e., unacknowledged expectancy violations) or from unsuccessful pursuit of conflict goals (Bevan et al., 2008). This is further clarified by the striking and theoretically meaningful divergence in the effects of actor conflict avoidance (an apparent protective factor) versus partner conflict avoidance (a risk factor) on recovery sabotage. Consistent with the demand–withdraw literature (e.g., Caughlin & Vangelisti, 2000), the link between partner conflict avoidance and later actor recovery sabotage may reflect a withdraw-demand sequence that plays out beyond the temporal boundaries of conflict. Conflict avoidance may protect against one’s own recovery sabotage in the short term. However, when partners are not complicit in avoiding conflict, the relatively fleeting protective effect for the conflict avoider comes at a cost to the partner, and possibly the relationship overtime, as opportunities for recovery are lost. These results highlight the cost of conflict suppression and the benefit of airing disagreements (even if they remain unresolved in the near-term) for conflict recovery and, potentially, for the long-term health of the relationship.

Recovery sabotage and conflict resolution

Finally, in this study, variation in recovery sabotage was not related to conflict resolution when other facets of conflict were controlled. These results suggest that recovery sabotage is not just a marker of unresolved conflict; some people who resolved conflict engaged in recovery sabotage and some people who did not resolve conflict were able to refrain from recovery sabotage. Moreover, partner recovery sabotage and conflict resolution were differentially associated with satisfaction and relationship stability 1 year later. These findings further validate the independence of recovery sabotage from conflict resolution and highlight the dyadic context in which recovery sabotage impacts a key relationship outcome.

Implications

One of the primary goals of this study was to identify for the first time how post-conflict behavior is associated with several interrelated processes that influence the course of romantic conflict: both partners’ attachment histories, affect and behavioral strategies employed during conflict, and conflict resolution. Results of this study document that, within this set of interrelated processes, there are distinct patterns of independent influence on post-conflict behavior. Specifically, the direct effects of attachment on recovery sabotage were not reduced when models included conflict affect, behavior, and resolution. Likewise, the direct effects of actor and partner conflict avoidance on recovery sabotage were independent of attachment, affect, and conflict resolution. Additionally, the direct effects of conflict resolution and recovery sabotage on longitudinal outcomes were independent of each other. These findings contribute important new evidence that post-conflict behavior is multiply determined and plays a distinct role in the course of relationships overtime.

Caveats and limitations

Despite its modest sample size, this study has several strengths, including the diversity of the sample and multilevel longitudinal design. However, results from this community sample may not generalize to clinically distressed couples, in which conflict and recovery processes may be compromised; this hypothesis should be tested empirically. As a reviewer pointed out, it is possible that partners may have disagreed about whether the relationship had dissolved at Time 2 and that our use of stability data for couples in which only one partner responded (n = 38) privileged the respondent’s perspective over the non-respondent’s. Finally, although recovery sabotage is a compelling indicator of conflict recovery, it is not the only possible marker of poor recovery. We expect that poor recovery is marked by the presence of negative behavior and affect as well as the absence of positive behavior and affect. These possibilities should be investigated in future research on conflict recovery.

Conclusion

Nearly four decades have passed since Gottman and colleagues began to publish landmark research demonstrating that how couples engage in conflict is associated with relationship satisfaction and stability (Gottman, 1979). This study integrates the long-standing romantic conflict literature and the more recently introduced concept of conflict recovery to document that recovery sabotage is a distinct relationship process tied to a key aspect of developmental history and behavioral interactions between partners during conflict. In particular, partners of conflict avoiders were more likely to sabotage recovery and partners of recovery saboteurs were less satisfied 1 year later, illustrating the pernicious process through which conflict avoidance undermines relationships over time. Partners’ fates were further entwined: each person’s satisfaction was associated more strongly with the other’s recovery sabotage behavior than with their own. Consistent with recent evidence that indirect negative communication during conflict can be more damaging than direct negative tactics and that direct negative tactics sometimes have positive consequences (Overall & McNulty, 2017), our findings suggest that partners who directly engage in conflict (rather than avoiding conflict) may be best positioned to set conflict aside when needed in order to pursue other relationship goals and promote each other’s satisfaction over the long-term. These findings have important implications for how couples solve everyday problems when situational demands (e.g., work schedules, parenting responsibilities) interrupt conflicts in progress and require effective recovery to pursue other goals even when conflicts are not yet resolved. While many therapeutic interventions target conflict behavior and affect management to promote relationship satisfaction, our findings suggest that post-conflict behavior may also be a fruitful avenue for intervention.

Footnotes

Authors’ note

Katherine C. Haydon, Cassandra Jonestrask, and Haley Guhn-Knight belong to the Department of Psychology and Education, Mount Holyoke College; Jessica E. Salvatore belongs to the Department of Psychology and the Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University.

Acknowledgment

We thank Paula Pietromonaco for her comments.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding from Mount Holyoke College to Katherine C. Haydon supported this research. Jessica E. Salvatore is supported by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health (F32AA022269 and K01AA024152).

Notes

References

Berscheid

Ammazzalorso

(2001). Emotional experience in close relationships. In Fletcher

G. J. O.

Clark

M. S.

(Eds.), Blackwell Handbook of Social Psychology: Interpersonal Processes (pp. 308–330). Oxford: Blackwell Publishers.

Bevan

J. L.

(2014). Dyadic perceptions of goals, conflict strategies, and perceived resolvability in serial arguments. Journal of Social and Personal Relationships, 31, 773–795.

Bevan

J. L.

(2015). Introduction to the special issue on 30 years of research on serial argument in interpersonal relationships. Argumentation and Advocacy, 52, 1–5.

Bevan

J. L.

Finan

Kaminsky

(2008). Modeling serial arguments in close relationships: The serial argument process model. Human Communication Research, 34, 600–624.

Bevan

J. L.

Hefner

Love

(2014). An exploration of topics, conflict styles, and rumination in romantic nonserial and serial arguments. Southern Communication Journal, 79, 347–360.

Caughlin

J. P.

Vangelisti

A. L.

(2000). An individual difference explanation of why married couples engage in the demand/withdraw pattern of conflict. Journal of Social and Personal Relationships, 17, 523–551.

Collins

W. A.

Aguilar

Hennighausen

Hyson

Jimerson

Levy

… Sesma

(1999). Scales and coding manual for observed interactions in romantic relationships. Unpublished manuscript, University of Minnesota, Minneapolis, MN.

Cox

M. J.

(1991). Marital and parent-child relationships study. Unpublished manuscript, University of North Carolina, Chapel Hill, NC.

Ewart

C. K.

Taylor

C. B.

Kraemer

H. C.

Agras

W. S.

(1991). High blood pressure and marital discord: not being nasty matters more than being nice. Health Psychology, 10, 155–163.

10.

Feeney

J. A.

(2004). Hurt feelings in couple relationships: Towards integrative models of the negative effects of hurtful events. Journal of Social and Personal Relationships, 21, 487–508.

11.

Fraley

R. C.

Heffernan

M. E.

Vicary

A. M.

Brumbaugh

C. C.

(2011). The Experiences in close relationships—relationship structures questionnaire: A method for assessing attachment orientations across relationships. Psychological Assessment, 23, 615–625.

12.

Goldberg

A. E.

Garcia

(2015). Predictors of relationship dissolution in lesbian, gay, and heterosexual adoptive parents. Journal of Family Psychology, 29, 394–404.

13.

Gottman

J. M.

(1979). Detecting cyclicity in social interaction. Psychological Bulletin, 86, 338–348.

14.

Gottman

J. M.

(1993). The roles of conflict engagement, escalation, and avoidance in marital interaction: a longitudinal view of five types of couples. Journal of Consulting and Clinical Psychology, 61, 6–15.

15.

Gottman

J. M.

(2014). What predicts divorce? The relationship between marital processes and marital outcomes. New York, NY: Psychology Press.

16.

Gottman

J. M.

Levenson

R. W.

(1986). Assessing the role of emotion in marriage. Behavioral Assessment, 8, 31–48.

17.

Gottman

J. M.

Levenson

R. W.

(1992). Marital processes predictive of later dissolution: Behavior, physiology, and health. Journal of Personality and Social Psychology, 63, 221–233.

18.

Gottman

J.M.

Levenson

R.W.

(1999). Rebound from marital conflict and divorce prediction. Family Process, 38, 287–292.

19.

Overbeek

Lichtwarck-Aschoff

Engels

R. C. M. E.

(2013). Do conflict resolution and recovery predict the survival of adolescents’ romantic relationships? PloS one, 8, e61871.

20.

Hatfield

Cacioppo

J. T.

Rapson

R. L.

(1994). Emotional Contagion. Cambridge, UK: Cambridge University Press.

21.

Haydon

K. C.

Roisman

G. I.

Burt

K. B.

(2012). In search of security: The latent structure of the Adult Attachment Interview revisited. Development and Psychopathology, 24, 589–606.

22.

Hendrick

S. S.

(1988). A generic measure of relationship satisfaction. Journal of Marriage and the Family, 50, 93–98.

23.

Hojjat

(2000). Sex differences and perceptions of conflict in romantic relationships. Journal of Social and Personal Relationships, 17, 598–617.

24.

Karney

B. R.

Bradbury

T. N.

(1995). The longitudinal course of marital quality and stability: A review of theory, method, and research. Psychological Bulletin, 118, 3–34.

25.

Kenny

D. A.

(2015). An interactive tool for testing distinguishablity and nonindependence in dyadic data [Computer software]. Retrieved from https://davidakenny.shinyapps.io/Dingy/

26.

Kenny

D.A.

Kashy

D.A.

Cook

W.L.

(2006). The Analysis of Dyadic Data. New York, NY: Guilford Press.

27.

Mikulincer

Shaver

P. R.

Pereg

(2003). Attachment theory and affect regulation: The dynamics, development, and cognitive consequences of attachment-related strategies. Motivation and Emotion, 27, 77–102.

28.

Overall

N. C.

McNulty

J. K.

(2017). What type of communication during conflict is beneficial for intimate relationships? Current Opinion in Psychology, 13, 1–5.

29.

Prager

K. J.

Shirvani

Poucher

Cavallin

Truong

Garcia

J. J.

(2015). Recovery from conflict and revival of intimacy in cohabiting couples. Personal Relationships, 22, 308–334.

30.

Salvatore

J. E.

Kuo

S. I.

Steele

R. D.

Simpson

J. A.

Collins

W. A.

(2011). Recovering from conflict in romantic relationships: A developmental perspective. Psychological Science, 22, 376–383.

31.

Simpson

J. A.

Collins

W. A.

Tran

Haydon

K. C.

(2007). Attachment and the experience and expression of emotions in romantic relationships: A developmental perspective. Journal of Personality and Social Psychology, 92, 355–367.

32.

Whitton

S. W.

Waldinger

R. J.

Schulz

M. S.

Allen

J. P.

Crowell

J. A.

Hauser

S. T.

(2008). Prospective associations from family-of-origin interactions to adult marital interactions and relationship adjustment. Journal of Family Psychology, 22, 274–286.