Our selection of potential mediating mechanisms was guided by findings from a profusion of studies (e.g., Cook, Fine, & House, 1995; Mischel, 2004; Mischel & Shoda, 1995) that have investigated the underlying dispositions that determine how individuals interpret and respond to interpersonal situations. Researchers generally agree that three types of constructs are particularly important. First, a person’s perception of an interpersonal event or circumstance is influenced by their emotional state (Howard, 1995). Second, people tend to possess cognitive frameworks or schemas (Howard, 1995; Mischel, 2004) regarding the nature of the self, others, and relationships that influence how they interact with others. Finally, individuals have traits that are indicative of a characteristic style of interacting with others or of engaging the environment (Mischel, 2004). These three factors are thought to inform interpersonal interaction by influencing the stimuli or cues that are selected for attention, the significance ascribed to these stimuli, and the path of action chosen from the perceived alternatives. Thus we set out to select specific constructs that corresponded to each of these three fundamental dispositions.

Our selection was informed by three criteria. There had to be strong evidence that the construct is: (1) learned during childhood in the family of origin based on interactions with parents; (2) assumed to be relatively stable by mid-to-late adolescence; and (3) established as an influence on interaction with others, especially adult romantic partners. It seems reasonable that any construct meeting these criteria might at least partially mediate the impact of parenting on interaction with romantic partners. Based on findings from extant research we identified anger management as the emotional state, attachment style and hostile attribution bias as the schemas, and self-control as the interpersonal trait important for inclusion in the developmental model to be tested. Although a few studies have examined the mediating effect of attachment style, to our knowledge no research has investigated the extent to which these other variables mediate the link between parenting and interaction with romantic partners. In the sections that follow, we review the literature associated with each of these constructs. Next, we formulate hypotheses regarding the manner in which each mediates the association between the type of parenting one is exposed to as a youth and the style of interaction one engages in with an adult romantic partner.

While we argue that in large measure the effects of parenting on warm, loving interactions with a romantic partner is mediated by secure attachment and self-control and that the impact of harsh parenting is mediated by anger management, insecure attachment, hostile attribution bias, and self-control, it is unlikely that these four variables completely explain the effect of parenting on relationships with romantic partners. Rather, we expect that both supportive and harsh parenting will continue to be associated with behavior displayed toward romantic partners after taking into account the posited mediating processes.

This expectation rests on the assumption that parents influence their offspring’s relations with intimate partners, at least in part, through their interactions during childhood (Bryant & Conger, 2002; Conger et al., 2000, Simons et al 2012). Parental treatment of the child provides lessons regarding behavior that is suitable and appropriate between individuals who love one another (Straus & Gelles, 1990, Simons et al., 2012). Presumably, this modeling conveys behavioral scripts or social skills that are tacitly relied upon and enacted in interaction with romantic partners. To the extent that this is the case, one would expect a direct effect from supportive parenting to warm, loving behavior toward a romantic partner and a direct effect from harsh parenting to hostile, aggressive behavior toward a romantic partner.

Our model includes two control variables: family structure and gender. There is evidence that compared to never-divorced, two parent households, single and divorced parents tend to show more hostility and less warmth to their children (Simons & Associates, 1996). Further, parental relationship transitions that lead to changes in family structure are difficult for children. For example, subsequent divorces are more detrimental to offspring than the first divorce and parental remarriage is associated with outcomes as poor as, and in some cases worse than, youth living in single-parent households (Simons, Chen, Simons, Brody & Cutrona, 2006). Children of divorce experience various long-term negative consequences including less satisfying romantic relationships as adults, increased intimate partner violence, and a greater likelihood of their own marriage ending in divorce (Amato, 1999). The second control variable is gender. Parents tend to show more hostility to boys than to girls (Hagan & Kay, 1990) and men are more likely to display hostility and aggression than women (Simons, Simons & Wallace, 2004). Thus we include these two controls in order to eliminate the possibility that any relationships between our study constructs are distorted due to a common association with gender or family structure.

Bivariate correlations, means, and standard deviations for all study variables are presented in Table 1. The pattern of correlations is mostly consistent with the study hypotheses. Beginning with the parenting measures, both supportive and harsh parenting are correlated with warmth and hostility toward the romantic partner. Harsh parenting shows a significant association with all of the potential mediators, whereas secure attachment, self-control, and anger are significantly related to supportive parenting. Further, all of the proposed mediators are correlated with hostility toward the romantic partner, and all of them, with the exception of hostile attribution bias, are also associated with warmth toward the romantic partner. In addition, all of the potential mediators are significantly correlated with each other.

Given that MPlus allows imputation of missing data in SEM using the full maximum likelihood (FIML) method, the N for this modeling was slightly higher than that for the listwise correlations reported in Table 1. FIML is an unbiased approach that provides more power than the listwise procedure because all observed information is utilized in order to produce parameter estimations (Acock, 2005). It should be noted that we found no evidence that supportive and harsh parenting interacted to influence either the proposed mediators or outcomes. We began by running a fully recursive model controlling for gender and family instability. This model includes all possible pathways between the variables. Then, in an effort to obtain a more parsimonious model, we deleted all nonsignificant paths (t < 1.5) and ran the model again (e.g., reduced model). The difference in chi-square between the fully recursive and reduced models was not significant (Δχ²= 4.85, p= .68) indicating that the reduced model provided a more parsimonious fit of the data. Figure 1 displays the results of the reduced model for the total sample.

With few exceptions, the pattern of findings shown in Figure 1 is consistent with expectations. Beginning with the effect of parenting on the four proposed mediators, there are significant paths from both supportive and harsh parenting to self-control. Beta coefficients are .20 and −.23, respectively. Similarly, both types of parenting are related to secure attachment. Beta (β) is .30 for supportive parenting and −.18 for harsh parenting. However, only harsh parenting is related to hostile attribution bias (β=.25). Harsh parenting is the only parenting variable associated with low anger management (β =.19). Paths from supportive parenting to hostile attribution bias and anger management are not significant. The potential mediators show many of the anticipated associations with behavior toward romantic partners. Both self-control and secure attachment demonstrate small, albeit significant, associations with warmth toward romantic partner (β = .14 and .16, respectively). Low anger management shows a strong relation with hostility toward romantic partner whereas there is no significant association between anger management and warmth toward partner. In addition, Figure 1 shows that self-control, secure attachment, and hostile attribution bias are all related to low anger management. Thus these three constructs appear to have the expected indirect relation on hostility toward romantic partner through low anger management.

Next, we examined whether the paths from harsh parenting to the mediators were significantly stronger than paths from supportive parenting to the mediators. This was done by comparing the change in chi-square between models that constrained the paths from the two types of parenting to be the same to models that allowed the paths to differ. The difference in chi-square between the two models was significant (Δχ²= 9.09, p< .05) for hostile attribution bias indicating that harsh parenting has a significantly stronger impact than supportive parenting on hostile attribution bias. The same procedure was used to evaluate whether harsh parenting had a stronger impact than supportive parenting on self-control and secure attachment. Again, the difference in chi square was significant for both tests (Δχ²= 13.91, p< .05 and Δχ²= 16.99, p< .05, respectively), indicating that the effect of harsh parenting is greater than that of supportive parenting for self-control but significantly lower than that of supportive parenting for secure attachment. Harsh parenting also had a significantly stronger impact than supportive parenting on anger. Consistent with expectations, results indicated that the path from anger management to hostility toward partner is significantly larger (Δχ²= 11.55, p< .05) than the nonsignificant path (not depicted in the reduced model) from anger management to warmth toward romantic partner.

Finally, Figure 1 indicates that supportive parenting continues to be associated with warmth toward romantic partner (β=.30) after taking into account any indirect effects through the mediating variables and harsh parenting continues to be associated with hostility toward romantic partner (β=.27) after removing the portion of the association explained by the mediators. Results showed that the effect of supportive parenting on warmth to romantic partners had a significantly greater effect than harsh parenting had on hostility toward romantic partners (Δχ²= 11.52, p< .05).

Table 2 presents the results of using the delta method to test the significance of the various indirect effects suggested in Figure 1. The table indicates that all of the indirect effects depicted in Figure 1 are significant (p< .05). Both supportive and harsh parenting influence warmth toward romantic partner through secure attachment and self-control. Further, supportive and harsh parenting affect low anger management through secure attachment and self-control, while harsh parenting impacts low anger management through hostile attribution bias. In turn, secure attachment, self-control, and hostile attribution bias influence hostility toward romantic partner through low anger management.

Summarizing, the tests of the indirect effects indicate that parental behavior has more of an influence on the hostility than the warmth that adult children display toward their romantic partners. Further, the findings suggest that harsh parenting is more consequential than supportive parenting in terms of predicting hostility toward the romantic partner. Finally, the indirect effect of harsh parenting supports a causal sequence where harsh parenting has both a direct effect on anger management, as well as an indirect effect through its impact on hostile attribution bias and self-control, and low anger management, in turn, increases the probability of hostility toward the romantic partner. When the direct and indirect effects are summed, the model explains 12 percent of the variation in warm, loving behavior and 26 percent of the variation in hostile, aggressive behavior toward romantic partners.

Previous research has established a link between the parenting practices to which children are exposed in their family of origin and their patterns of interaction with adult romantic partners (Bryant & Conger, 2002; Conger, et al., 2000). The present study extended this research in three respects. First, in contrast to earlier studies, the current study investigated the independent effects of supportive parenting compared to harsh parenting. We used the same approach for romantic relationships by assessing the precursors of warm/loving compared to hostile/aggressive interactions. This approach allowed us to trace pathways from positive and negative starting points in the family of origin to positive and negative outcomes in young adult relationships. Second, although researchers have speculated about the social and psychological factors that likely mediate the impact of parenting practices on behavior toward romantic partners, few studies have explicitly specified those factors and, to our knowledge, none have tested them empirically. The current study identified anger management, attachment style, hostile attribution bias, and self-control as important constructs for investigation. Third, unlike earlier studies, ours uses prospective, longitudinal data from an African American sample. Findings provided strong support for a developmental model where parenting shapes youths’ interactional styles, schemas and traits, which in turn influence quality of interaction with romantic partners. Next, we summarize and interpret the findings.

First, the constructs that link parenting to warm interactions with romantic partners are somewhat different from those that link parenting to hostile interactions with romantic partners. As expected, we found that self-control, attachment style, and hostile attribution bias partially mediated the impact of parenting on hostility toward romantic partner. Furthermore, the effects of self-control, attachment style, and hostile attribution bias were mediated by anger management. This pattern of findings is consistent with the arguments of self-control theory (Burt et al., 2006), attachment theory (Simpson et al., 1996) and Dodge’s theory of hostile attribution bias (Holzworth-Munroe, 2000). The results suggest that individuals who lack patience and distrust others are prone to angry outbursts that often involve aggression toward romantic partners.

Next, in addition to its indirect effect on anger management through self-control, attachment style, and hostile attribution bias, harsh parenting also showed a direct effect on this variable. This finding is consistent with the argument that harsh, explosive parenting teaches children that strong negative emotions often cannot or need not be controlled (Gottman et al., 1997). We investigated whether warm, nurturant parenting would directly influence anger management, but this was not the case. Rather, the impact of such parenting practices on anger management was limited to its indirect effect through attachment style and self-control. This finding is consistent with the predictions of attachment theory (Mickulincer & Shaver, 2007; Thompson, Lewis, & Calkins, 2008) and self-control theory (Burt et al., 2006).

The findings regarding the determinants of warmth toward romantic partner were quite different than those for hostility toward romantic partner. Although attachment style and self-control partially mediated the effect of parenting on warmth toward romantic partner, this was not the case for hostile attribution bias. Further, as expected, the effects of attachment and self-control on warmth toward partner were not mediated by anger management. Indeed, anger management was not related to warmth toward partner. This suggests that individuals who have trouble controlling their anger can be quite warm as long as they are not frustrated or provoked (Ellis & Malamuth, 2000; Simons, et al., 2012). Secure attachment is characterized by trust and the capacity for intimacy, and self-control entails empathy and the ability to delay gratification. Our findings suggest that these competencies directly increase the probability of warm and loving behaviors toward a romantic partner.

Importantly, given the study’s focus on mediation, we found that both supportive and harsh parenting continued to show direct effects on behavior toward romantic partners after taking into account the impact of the mediators. This finding is consistent with the argument that parents influence their offspring’s relations with intimate partners, at least in part, through the behavior that they model in interaction with them during childhood (Bryant, 2006; Bryant & Conger, 2002; Conger et al., 2000). Children appear to acquire scripts or skills that are tacitly relied upon and enacted during interaction with romantic partners. Future research should endeavor to identify the nature of these scripts and skills.

The FACHS data used in the present study afforded several advantages such as longitudinal data, the separate assessment of supportive and harsh parenting, repeated measures of parenting behaviors during late childhood and early adolescence, assessment of the mediating variables during late adolescence and assessment of adult romantic relationships during early adulthood. Further, we used an African American sample to address the research questions while past research has focused primarily on European American samples. However, there were also limitations inherent in the data that need to be mentioned. First, by using an African Americans sample, the homogeneity may preclude generalization of the findings to other ethnic groups. Future research is needed in order to replicate our findings with more diverse samples.

Second, there are limitations associated with some of the measures. For example, our investigation of processes in the family of origin focused exclusively on parenting practices by the primary caregiver. Because less than half of the sample consisted of two-parent families, the study design did not include collecting data on parenting practices of the fathers or marital interactions between parents. However, some studies report that parenting practices serve as stronger predictors of adult intimate relationships than marital interactions (Bryant & Conger, 2002; Simons, Lin, & Gordon, 1998). Nevertheless, it would be worthwhile to investigate the extent to which the competencies that we found to mediate the impact of parenting practices also mediate any effect of marital interaction on an offspring’s romantic relationships. Further, we did not have data on the duration of the romantic relationships and our assessment of behavior toward romantic partners was obtained as the subjects were entering early adulthood. Therefore we have no way of knowing the extent to which interaction in their current dating relationships portend quality of the more long-term relationships that will likely be established in the coming years. There is evidence, however, that teen and early adult romantic relationships often serve as a training ground for later intimate relationships. Studies show that hostile and violent romantic relationships during late adolescence tend to predict similar relationship dynamics in adulthood (Conger et al., 2000). This suggests that those individuals in our sample who reported high hostility toward romantic partners in their current dating relationships are at risk for troubled, unstable adult romantic relationships. Future waves of data will allow us to examine these expectations. Finally, we relied on self-report data, which has the problem of method variance. However, this was diminished somewhat because the study variables were measured at different waves across several years.

Third, the respondents were 10 years of age when the study began. Presumably, parenting practices during early childhood have an impact on the development of the mediating variables in our study. It could be that the relationship between parenting practices and the mediators would be even stronger if we had measures of parenting at earlier ages. However, one’s general approach to parenting is somewhat stable (Simons, Su & Simons, 2013) so our measures are likely a good indicator of the quality of parenting over time. Of course, due to the bidirectional nature of the parent-child relationship, in the same way that earlier parenting contributes to later parenting and child cognitions/behaviors, it is also the case that earlier child cognitions/behaviors contribute to later parenting. Thus, because family influence is a reciprocal process, it could be that anger management, attachment style, hostile attribution bias, and low self-control form to some degree prior to adolescence and therefore influence the behavior of parents toward children at age 10. Our study doesn’t allow us to investigate this. However, theory (e.g. attachment) as well as past research (e.g. see Dodge’s work on hostile attribution bias) suggest that parenting is the more fundamental cause operating in this reciprocal relationship. Thus, it is likely the case that early parenting gives rise to the development of the mediating variables we address in the current study. Future research would benefit from examining these issues.

Despite these limitations, the current study might be seen as offering strong support for a developmental model of romantic relationships and providing important information regarding the avenues whereby experiences in the family of origin affect the quality of an individual’s adult intimate relationships. Overall, our findings indicate that the type of parenting received as a child influences adult romantic relationships through its impact on various competencies. These results point to the importance of future research utilizing a developmental perspective where childhood and adolescent experiences, especially in the family of origin, give rise to attitudes, traits and schemas that influence the way an individual relates to romantic partners.

Our findings suggest a number of implications regarding prevention and intervention. For example, parenting education classes could benefit by emphasizing the importance of supportive parent-child interactions for the long-term adjustment of their offspring including their future romantic relationships. Discussion of the importance of the family of origin in the development of emotional states, schemas, and interactional styles that may be detrimental to the quality of one’s intimate relationship may enhance the effectiveness of relationship or marriage education programs. Further, clinical intervention efforts with hostile romantic partners might be improved by focusing upon the development of anger management, as well as upon the underlying issues of distrust and self-control that give rise to this emotional volatility. Additionally, efforts to enhance trust and self-control combined with attempts to reduce anger and hostility in a clinical setting can lead to enhanced warmth and support exchanged by the romantic partners. Verbal expressions of a positive nature are associated an increase in relationship satisfaction (Cohan & Bradbury, 1997) while negative premarital communication is associated with later marital distress and an increased risk of divorce (Clements, Stanley & Markman, 2004). The benefits of such efforts may go beyond individual well-being and relationship satisfaction. For example, there is evidence that reducing hostility and increasing relationship quality is also associated with a reduction in health risks (Guyll, Cutrona, Burzette, & Russell, 2010). Thus, efforts to reduce relationship discord and increase relationship quality will benefit adult intimate partners, their children, and society.