Parental Modeling and Deidentification in Romantic Relationships Among Mexican-origin Youth

Abstract

This study investigated youth’s modeling of and de-identification from parents in romantic relationships, using two phases of data from adolescent siblings, mothers, and fathers in 246 Mexican-origin families. Each parent reported his/her marital satisfaction and conflict, and youth reported on parent-adolescent warmth and conflict at Time 1. Youth’s reports of modeling of and de-identification from their mothers and fathers and three romantic relationship outcomes were assessed at Time 2. Findings revealed that higher parental marital satisfaction, lower marital conflict, and higher warmth and lower conflict in parent-adolescent relationships were associated with more modeling and less de-identification from parents. Moreover, higher de-identification was linked to a greater likelihood of youth being involved in a romantic relationship and cohabitation, whereas more modeling was linked to a lower likelihood of cohabitation and older age of first sex. Discussion underscores the importance of assessing parental modeling and de-identification and understanding correlates of these processes.

Forming and maintaining positive romantic relationships are salient developmental tasks of adolescence and young adulthood (Collins, Welsh, & Furman, 2009; Masten, Burt, & Coatsworth, 2006). Such relationships have implications for psychosocial adjustment and physical health across adulthood (Diener & Seligman, 2004; Reis, Collins, & Berscheid, 2000). Developmental and family theorists have long postulated that romantic relationship characteristics have roots in early family experiences (Bryant & Conger, 2002), and a growing literature has documented intergenerational links between parents’ marital and parent-child relationship functioning and youth’s romantic relationship experiences (Conger, Cui, Bryant, & Elder, 2000). To contribute to this literature, this study examined whether youth learn from their parents’ examples, specifically, by treating their parents as role models, or conversely, treating them as foils and differentiating or de-identifying from them (Rhoades, Standley, Markman & Ragan, 2012). Focusing on Mexican-origin adolescent siblings from two-parent families with long-term couple relationships, we assessed the correlates of youth’s reports of modeling of and de-identification from mothers and fathers in the domain of romantic relationships.

Intergenerational Transmission of Romantic Relationships

Intergenerational similarities with respect to behaviors and attitudes have been a central focus of developmental and family research (Grusec & Hastings, 2007). These lines of study are guided by the notion that attitudes and behaviors are often transmitted from parents to children through social learning processes (Bandura, 1977). Evidence reveals positive associations between parents’ romantic relationship status and experiences and their offspring’s romantic relationships in adolescence and adulthood. For example, parental divorce increases the likelihood of offspring relationship dissolution (e.g., Amato & Booth, 1997), and interparental conflict is associated with offspring’s relationship conflict (e.g., Kinsfogel & Grych, 2004). Further, meta-analysis indicates a positive association between growing up in a violent home and being involved in subsequent marital violence (Stith et al., 2000). Even though most studies of intergenerational transmission of romantic relationship experiences are grounded in social learning theory (Bandura, 1977), measurement of parental modeling is rare (except, see Rhoades et al., 2012). Instead, investigators have invoked social learning processes as post-hoc explanations for similarities across generations. Further, the implicit assumption has been that parents’ socialization role is exclusively to foster their children’s similarity to themselves. Although prior findings reveal that youth resemble their parents in romantic relationship outcomes (e.g., Amato & Booth, 1997; Kinsfogel & Grych, 2004), the associations are modest (Stith et al., 2000). Such a pattern implies that socialization influences beyond modeling may come into play, including youth’s efforts to learn from their parents’ mistakes and take a different course in their own lives. Thus, in addition to measuring modeling, we also assessed youth’s reports of their efforts to de-identify or differentiate from their parents in the romantic relationship domain.

Another gap in knowledge is the correlates of romantic relationship experiences in ethnic minority families given that most research focuses on European American families. A tenet of cultural-ecological frameworks is that development is embedded within a larger sociocultural context, and that studies of majority youth may not generalize to youth from other cultural backgrounds (García Coll et al., 1996). A focus on Mexican-origin families is warranted because of the centrality of family relationships and an emphasis on interdependence among family members in Mexican culture (Cauce & Domenich-Rodríguez, 2002). Accordingly, we examined youth’s reports of modeling and de-identification in reference to their parents’ romantic relationships among Mexican-origin youth, the largest, youngest, and fastest growing Latino subgroup in the U.S. (U.S. Census Bureau, 2014). Mexican-origin/Latino youth also are a group that is important for the study of romantic relationships because they have the youngest median age of first marriage in the U.S. (Payne, 2015). Thus, a better understanding of the role of Mexican-origin parents in youth’s romantic relationship outcomes may inform efforts to promote this group’s romantic relationship functioning. Toward this end, our study goals were: (a) to identify the predictors of modeling and de-identification processes, including parents’ marital quality and characteristics of parent-youth relationships; and (b) to explore the links between modeling and de-identification processes and three domains of romantic relationship behaviors: relationship involvement, cohabitation, and age at first sex.

Modeling and De-identification from Parents

According to social learning theories, behaviors are often learned by observing and imitating others (Bandura, 1977). Youth may emulate the behaviors and attitudes that they observe in their parents’ romantic relationships, as well as those they experience in their own relationships with parents; furthermore, they may use those experiences to differing degrees in their own romantic relationships. Originating in psychoanalytic thought (Alfred Adler; see Ansbacher & Ansbacher, 1956) and in the sibling influence literature (McHale, Updegraff, & Whiteman, 2012; Whiteman, McHale, & Crouter, 2007), de-identification, in contrast, refers to the tendency for youth to define themselves as different and behave differently from their parents—that is, to treat their parents as foils rather than models. This suggests that children may learn from the experiences of their parents and define themselves as different from them. The concepts of modeling and de-identification are related to modeling and antimodeling in the sociology literature, with the notion that antimodeling refers to persons who exhibit behaviors and characteristics that individuals work to avoid (Francis, 1963). For instance, youth who observe that their parents are unhappy in their marriage may decide to adopt a different approach in their own romantic relationships. To date, research on de-identification processes in parent-youth relationships is limited, but preliminary work reveals some evidence of de-identification (Perez-Brena, Updegraff, & Umaña-Taylor, 2014). Qualitative data also showed that men become involved in parenting both because their own father was involved—or because their father was not involved, and they want to be different (Lamb & Tamis-Lemonda, 2004). Such findings suggest that men may choose to emulate their fathers—or to compensate by setting different goals. Results such as these underscore the importance of systematically measuring parental influence processes directly. In a handful of studies, both modeling and de-identification were measured via youth reports, and findings revealed that these processes were negatively correlated, but not opposite ends of the same continuum in sibling (Whiteman et al., 2007) and parent-child relationships (Perez-Brena et al., 2014). Trying to be different (de-identification)—treating parents as foils—is not the same as failing to model one’s parents (low modeling). Low levels of modeling may reflect separation and individuation processes as youth choose models beyond the family, but efforts to de-identify tend to occur in relationships marked by rivalry and conflict (in early writings of Alfred Adler; see Ansbacher & Ansbacher, 1956).

Rhoades and colleagues (2012) provided initial data on parents as role models in romantic relationships by examining adolescents’ ratings of the extent to which they considered their parents’ relationship as a model of a good relationship. Findings revealed that the association between parents’ marital status (divorced vs. married) and offspring’s poorer relationship adjustment was mediated by adolescents’ reports of whether their parents’ romantic relationships were models of healthy relationships. The investigators concluded that more detailed measurement of modeling was an important direction for research. In the current study, we measured youth’s reports of modeling and de-identification in the domain of romantic relationships as two distinct processes in an effort to illuminate mechanisms of intergenerational transmission. Grounded in family systems theory, which posits that individuals and subsystems within the family are interdependent and influence one another (Cox & Paley, 2003), and a body of research highlighting the distinct roles of mothers and fathers in Mexican American families (Cauce & Domenich-Rodríguez, 2002), we also measured modeling and de-identification from mothers and fathers to understand whether and how experiences with one parent cross over to have implications for experiences with the other (e.g., whether mother-son relationship qualities predicts modeling of fathers and vice versa).

Predictors of Parental Modeling and De-identification Processes

As noted, our first goal was to assess whether parents’ marital and parent-adolescent relationship quality predicted youth’s modeling of and de-identification from parents five years later. We expected, first, that marital quality would be positively related to youth’s reports of parental modeling, a hypothesis grounded in the literature on intergenerational transmission of marriage. Adult children with married parents were more likely to rate their parents’ relationship as a good model for their own relationships as compared to those with divorced or never married parents (Rhoades et al., 2012). Going beyond marital status, research also has documented that parents’ marital quality contributes to the extent of similarity between parents’ and their offspring’s marriage-related attitudes (Cunningham & Thornton, 2006). Together, these findings suggest that parents whose marital relationships are characterized by positive qualities may be more attractive models for youth and that their children are more likely to see them as models for their own romantic relationships. Based on this line of study, we expected that youth with parents who reported higher levels of marital satisfaction and less marital conflict would be more likely to see their parents as role models. Although de-identification from parents has not been studied as a distinct process (except, see Perez-Brena et al., 2014), we expected that youth whose parents described more marital conflict would report more efforts to de-identify from their parents in this domain.

We also expected that youth’s modeling of and de-identification from parents would be linked to parent-youth relationship qualities (Bandura, 1977; Darling & Steinberg, 1993). According to social learning theory, individuals are more likely to imitate warm and nurturant models (Bandura, 1977), and empirical work suggests that parental warmth and acceptance occur within the context of supportive parenting (Darling & Steinberg, 1993). Youth who lack support or experience negativity in their relationships with parents, in contrast, may shift their focus to others (e.g., siblings, friends) as sources of understanding and support and even role models (Goldscheider & Goldscheider, 1998). Conflict may be particularly salient in immigrant families as parents and youth negotiate changes in their relationships in response to both the developmental changes of adolescence and parent-youth differences in acculturation (Telzer, 2010). Moreover, given that Mexican American cultural values emphasize family cohesiveness (Cauce & Domenech-Rodríguez, 2002), parent-youth relationships may be especially influential in youth’s decisions to model or de-identify from their parents. For these reasons, we expected that closer and less conflictual parent-youth relationships would predict more modeling of and less de-identification from parents.

Because social learning theories also assert that similarity between observers and models promotes modeling (Bandura, 1977), the gender constellation of the parent-youth dyad may play a role in intergenerational transmission processes. As youth tend to identify more with same-gender models (McHale, Updegraff, Helms-Erikson, & Crouter, 2001), we hypothesized that they would be more likely to model their same-gender parent. The sibling literature suggests, in contrast, that youth are more likely to de-identify with individuals who have similar objective characteristics as a means of reducing competition and rivalry and thus that de-identification also will be more evident in same-gender dyads (McHale et al., 2012). However, given the lack of theory and empirical research on de-identification from parents, we did not advance a specific hypothesis regarding the moderating role of parent-youth gender constellation in de-identification.

Linking Modeling of and De-identification from Parents to Youth’s Relationship Outcomes

Our second goal was to link modeling of and de-identification from parents to three indicators of youth romantic relationship outcomes: relationship involvement, cohabitation, and age at first sex. Romantic relationship involvement is normative in adolescence and young adulthood, (Collins et al., 2009), and cohabitation is becoming more common in young adulthood (Rose-Greenland & Smock, 2013), Some work shows, however, that cohabitation at younger ages is associated with divorce/break up (Kuperberg, 2014). More generally, older ages of first sex and cohabitation are beneficial for healthy adjustment (e.g., Kuperberg, 2014). Given that youth in this study came from two-parent families with long-term couple relationships characterized by high marital satisfaction and low conflict (Wheeler, Updegraff, & Thayer, 2010)—setting an example for relationship harmony and stability—we expected that more modeling and less de-identification from parents’ romantic relationship experiences would be positively linked to relationship involvement and negatively linked to cohabitation and age of sexual debut in later adolescence/early adulthood.

The Current Study

In sum, using data from a longitudinal study of Mexican-origin parents and both younger (adolescents) and older siblings (young adults), we took the novel step of assessing youth’s reports of the extent to which they tried to be like (model) and be different from (de-identify) their mothers and fathers in the domain of romantic relationships. The first goal was to examine whether youth’s modeling and de-identification were predicted by their family experiences five years earlier. Specifically, we expected that parents’ marital (more satisfaction, less conflict) and parent-youth relationship quality (more warmth, less conflict) would predict more modeling of and less de-identification from parents. We also tested whether modeling and de-identification were more evident in same-gender versus mixed-gender parent-youth dyads. In addition, birth order (younger vs. older sibling) was included in all models given potential birth order differences, as there is some evidence indicating that parental influences are more important for older as compared to younger siblings (McHale et al., 2001). The second goal was to investigate how modeling of and de-identification from parents were linked to youth’s concurrent romantic relationship outcomes. Given that youth in this sample grew up in long-term married families, we predicted that modeling would be positively related and de-identification would be negatively related to more healthful romantic relationship outcomes, including higher relationship involvement, lower likelihood of cohabitation, and older age at first sex. In all analyses we included control variables that have been linked to marital quality (i.e., parents’ marital status, relationship length), and socioeconomic status indicators (Bradbury, Fincham, & Beach, 2000).

Method

Participants

We used data from a longitudinal study of gender and family socialization processes in 246 Mexican-origin families (Updegraff, McHale, Whiteman, Thayer, & Delgado, 2005). The participating families were recruited through schools in a southwestern metropolitan area. Given the larger goal of the longitudinal study was to examine normative family processes in Mexican-origin families with adolescents, to be eligible for the study, at Time 1 (T1) families included a biological mother, biological/adoptive father (minimum of 10 years), younger sibling, and older sibling, all living in the home and willing to participate. Further, mothers had to self-identify as being of Mexican origin, and fathers had to be working at least 20 hours/week. Although not an eligibility criterion, 93% of fathers also were of Mexican descent. We focused on two-parent families with employed fathers given our interests in the roles of both mothers and fathers in family and gender dynamics. T1 (in 2002–2003) data collection occurred when younger siblings were in 7^th grade (M = 12.51 years; SD = .58) and their older siblings averaged 15.48 years old (SD = 1.57). Time 2 (T2) data were collected approximately 5 years later, when younger siblings and older siblings averaged 18.18 and 21.11 years of age, respectively.

At T1, families ranged in SES; the percentage of families meeting federal poverty guidelines (18.3%) was similar to the percent in poverty for two-parent Mexican American families in the county from which the sample was drawn (18.6%; U.S. Census, 2000). Median household income was $40,000 (range from $3,000 to over $250,000). Most parents were born in Mexico (71% of mothers and 69% of fathers) and preferred to be interviewed in Spanish (66% of mothers, 67% of fathers). The majority of parents were married (92.7%; M duration = 19.33 years, SD = 5.00) and the remaining were in long-term unions (M duration = 17.00 years, SD = 4.56). Parents reported an average of 10 years of education (M = 10.33; SD = 3.73 for mothers, and M = 9.87, SD = 4.37 for fathers). Approximately 51% of younger siblings (n = 125) and 50% of older siblings (n = 123) were female. Youth were most likely to have been born in the U.S. (62%), and most completed the interview in English (83%).

To evaluate sample representativeness, we compared our sample to Mexican-headed two-parent families in the county from which it was drawn (U.S. Census Bureau, 2000). Similar percentages of families met federal poverty guidelines (18.3% versus 18.6%) and similar percentages of mothers (19.1%) and fathers (22.4%) completed high school relative to Mexican female (22.1%) and male adults (20.3%) in the county. Median household income, however, was slightly higher in our two-parent sample ($41,000) relative to the county ($32,000).

At T2, over 75% of the families participated (n = 184). Comparisons of the non-participant (n = 62) with the participant families (n = 184) revealed that nonparticipants had lower incomes (M = $36,941; SD =$28,584 versus M = $58,657; SD = $48,636 for participant families), t (244) = −3.33, p < .001), and lower maternal education (M = 9.48; SD = 3.45 versus M = 10.62; SD = 3.79 for participant families), t (244) = −2.09, p < .05 at T1. Thus, we controlled for T1 family SES in all analyses to improve estimation of missing data.

Procedures

At both time points, families participated in structured in-home interviews lasting two to three hours. Interviews were conducted separately with each family member. Bilingual interviewers read the questions aloud and entered responses in laptop computers. Families received honorariums ($100 at T1, $125 at T2) for participation. The University’s Institutional Review Board approved all procedures.

Measures

All measures were forward and back-translated into Spanish for local Mexican dialect by two bilingual research assistants. Final translations were reviewed by a third translator, and discrepancies were resolved by the research team (Knight, Roosa, & Umaña-Taylor, 2009).

Modeling and de-identification with parents in romantic relationships

Youth reported on their modeling and de-identification with their mothers’ and fathers’ romantic relationships at separate points in the T2 interview. We adapted Whiteman et al.’s (2007) measure of sibling modeling and de-identification to examine the degree to which youth looked to parents as a model or de-identified from their parents in the domain of romantic relationships (see Perez-Brena et al., 2014 for psychometrics). Using a scale from 1 (never) to 5 (very often) and a time frame of the past 12 months, youth rated four items to assess modeling (e.g., “My mother/father is a role model for how I should act in romantic relationships,” “My mother/father gives me advice about how to behave in romantic relationships”). Four additional items assessed adolescents’ views of parents as negative referents from whom to de-identify (e.g., “My mother/father’s romantic relationship is not a good example for me,” “I try to have a relationship that is different from my mother’s and father’s relationship.”) Cronbach’s αs ranged from .74 to .87 for older and younger siblings’ reports of modeling and de-identifying from mothers and fathers. Low correlations, r = −.12 to r = −.27, emerged between the modeling and de-identification indices, consistent with prior work (Perez-Brena et al., 2014; Whiteman et al., 2007), suggesting that modeling and de-identification processes are related but distinct. Correlations between parental warmth and modeling of parents ranged from r = .28 to r = .39, whereas parental warmth and de-identification correlations ranged from r = −.18 to r = −.30, suggestive of discriminant validity.

Marital relationship quality

Mothers and fathers reported on their marital satisfaction and conflict at T1 in the past 12 months. Parents rated marital satisfaction in 16 domains (e.g., communication, division of housework) on a 9-point scale (1 = extremely dissatisfied, 9 = extremely satisfied) using a measure adapted from Huston, McHale, and Crouter (1986). Cronbach alphas were .95 for mothers and .92 for fathers. Marital conflict was measured using the 5-item scale from Braiker and Kelley’s (1979) Relationship Questionnaire. Parents used a 9-point scale to rate items such as “How often do you feel angry or resentful towards your spouse?” Reliabilities were α = .67 and α = .68 for mothers and fathers.

Parent-youth relationship quality

Youth reported on their conflict and warmth with both mothers and fathers at T1. They rated the frequency of conflict with mothers and fathers during the past year (ranging from 1 = Not at all to 6 = Several times a day) in 12 domains (e.g., chores, bedtime/curfew, family obligations) using a measure developed by Smetana adapted for Mexican-origin individuals (Updegraff, Delgado, & Wheeler, 2009). Cronbach’s αs ranged from .81 to .86 for older and younger siblings’ reports of conflicts with mothers and fathers. Youth also rated maternal and paternal warmth using the 8-item Children’s Report of Parental Behavior Inventory (Schwarz, Barton-Henry, & Pruzinsky, 1985). Items were rated on a 5-point scale (1 = almost never to 5 = almost always) for the time period of the past year. Cronbach’s αs ranged from .87 to .91 for older and younger siblings’ reports of warmth with mothers and fathers

Youth’s romantic relationship outcomes

Youth reported on three romantic relationship outcomes at T2. Involvement was indexed by youth’s report of whether they were involved in a romantic relationship with a boyfriend or girlfriend for at least the past month (0 = no, 1 = yes). Cohabitation was indexed by youth’s reports of whether they were currently living with their partner in the same house or apartment (0 = no, 1 = yes). Lastly, youth reported on how old they were the first time they had sexual intercourse.

Family background characteristics

Youth reported their gender (0 = female, 1 = male) and age, and parents reported on their own educational levels and their annual incomes at T1. Family SES was measured with the log of household income (to correct for skewness), and mothers’ and fathers’ education level, which were standardized and averaged such that higher scores indicated higher SES (α = .78). Parents also reported on their marital status (0 = cohabitating, not legally married, 1 = legally married) and length of their relationship at T1. Nativity was reported by youth (1 = U.S.-born and 0 = Mexico-born).

Results

We used a combination of path analysis and logistic regression analysis to examine mothers’ and fathers’ marital quality (i.e., satisfaction, conflict) and parent-youth relationship quality (i.e., warmth, conflict) at T1 as predictors of youth’s T2 modeling and de-identification with mothers’ and fathers’ romantic relationships (five years later). In the same model, we assessed the links between youth’s modeling and de-identification with mothers and fathers and youth’s T2 relationship outcomes (i.e., involvement, cohabitation, age at first sex). Two sets of models, stacked by sibling, were estimated; one included modeling of and one including de-identification from both mothers and fathers using Mplus 7.11. Analyses used the full sample, and missing data were accounted for using full information maximum likelihood robust (FIML-R). The dependent variable residuals were correlated. To adjust for the non-independence of observations (data from two siblings and two parents clustered in families), family was used as a cluster variable in all analyses such that standard errors were adjusted. All continuous independent variables were grand mean centered. Family SES, parents’ relationship duration and marital status, youth gender, and sibling birth order were included in all models as controls.

To examine the moderating role of parent gender, we estimated a series of models in which the predictor coefficients were constrained to equality across mothers and fathers (i.e., constraining mothers’ and fathers’ parameters to be equal). The constrained models were compared to the fully unconstrained models using a log likelihood difference test. If the change in the log likelihood was significant, p < .05, we interpreted the findings as indicative of moderation (i.e., differences in patterns of findings by parent gender; Kline, 1998). If the two models did not result in significant change in the log likelihood, the constrained model was retained for all further analyses of the model. To examine the moderating role of youth gender and sibling birth order, we included interaction terms (e.g., maternal modeling × youth gender). The final models include only significant interaction terms as retaining interactions that are not significant contributes to an increase in standard errors (Aiken & West, 1991). We conducted follow up analyses to interpret significant interactions as outlined by Aiken and West (1991), including plotting and testing for significant simple slopes by group. Moderation by youth gender and birth order is described when significant. Moderation by nativity was examined but was never significant; thus, it was removed from the final models.

Descriptive Analyses

Table 1 provides bivariate relations between study variables, means, and standard deviations. Parents’ ratings of their marital satisfaction averaged above, and their ratings of marital conflict averaged below, the midpoints of the scales. Youth’s ratings of modeling and de-identification averaged around the midpoint on the rating scales, evidence of moderate degrees of both modeling and de-identification. To provide descriptive information on modeling and de-identification, we conducted a series of 2 (gender) × 2 (sibling: older vs younger) × 2 (parent: mother vs father) mixed model ANOVAs, with sibling and parent as within-group factors and modeling and de-identification as dependent variables. For modeling, findings revealed a parent effect, which was qualified by a parent × gender interaction, F (1, 278) = 22.88, p < .01; follow-ups revealed mean level differences in maternal modeling, such that daughters reported higher levels (M = 3.19, SD = 1.15) than sons (M = 2.67, SD = 1.02), F (1, 291) = 15.79, p < .01, but no differences in paternal modeling (M = 2.56; SD = 1.04 for daughters; M = 2.52; SD = 1.00 for sons). No significant effects emerged for de-identification.

Table 1.

Correlations and Descriptive Statistics for Study Variables (N = 246 families)

Variables	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19
Time 1
1. Youth gender	-	−.09	.05	.06	−.04	−.02	.03	.07	−.05	.05	.08	.16*	−.31*	−.01	−.01	.05	−.14†	−.16*	−.07
2. SES	−.02	-	−.04	.03	.13*	−.06	.03	.14*	.09	.09	−.15*	−.22*	.05	−.04	−.09	−.06	−.02	−.05	.16
3. Marriage duration	−.02	−.04	-	.11	−.02	−.01	−.08	−.10†	.04	.10	−.08	−.09	−.01	.09	.01	−.02	.07	−.06	.07
4. Marital status	.06	.03	.11†	-	.00	.05	−.07	.06	.02	−.03	−.06	−.04	.13	.06	−.06	.09	−.04	−.06	−.13
5. Marital sat (M)	−.04	.13*	−.03	.00	-	.34*	−.46*	−.20*	.08	.29*	−.17*	−.17*	.14†	.27*	−.30*	−.29*	−.21*	−.03	−.08
6. Marital sat (F)	.05	−.06	−.01	.05	.34*	-	−.22*	−.11†	.21*	.20*	−.09	−.04	.14†	.28*	−.29*	−.21*	−.04	.04	−.01
7. Marital con (M)	−.03	.03	−.08	−.07	−.46*	−.22*	-	.33*	−.11	−.20*	.19*	.16*	−.10	−.26*	.30*	.22*	.10	.03	−.10
8. Marital con (F)	−.10	.14*	−.10	.06	−.19*	−.12†	.32*	-	−.09	−.07	.04	−.03	−.01	−.13	.16†	.05	−.02	.02	−.02
9. Warmth (M)	−.06	.13*	.05	.07	.16*	.20*	−.11†	−.01	-	.43*	−.24*	−.13*	.37*	.28*	−.25*	−.14	−.17*	−.11	.22
10. Warmth (F)	.04	.18*	−.01	.02	.39*	.25*	−.20*	−.13*	.49*	-	−.15*	−.16*	.20*	.30*	−.23*	−.18*	−.14*	−.10	−.04
11. Conflict (M)	.09	−.02	−.07	.01	−.06	−.03	.08	−.02	−.41*	−.18*	-	.75*	−.05	.03	.22*	.09	.21*	.18*	.05
12. Conflict (F)	.02	.01	−.05	.05	−.06	−.03	.03	−.03	−.27*	−.16*	.73*	-	−.05	−.03	.26*	.15	.13†	.08	−.04
Time 2
13. Modeling (M)	−.15†	.06	−.09	−.14*	.10	.20*	−.22*	−.15†	.37*	.20*	−.14	−.18*	-	.62*	−.12	−.11	.02	.07	.33*
14. Modeling (F)	.07	.06	.13	−.12	.25*	.29*	−.31*	−.17*	.22*	.31*	−.12	−.15†	.57*	-	−.28*	−.27*	.06	−.01	.26*
15. De-id (M)	−.02	−.03	−.13	.00	−.29*	−.18*	.38*	.19*	−.29*	−.33*	.24*	.23*	−.22*	−.33*	-	.68*	.15*	−.08	.04
16. De-id (F)	.04	.02	−.15†	.03	−.17*	−.13	.26*	.14	−.25*	−.19†	.22*	.07	−.12	−.23*	.74*	-	.06	−.09	.12
17. Rel involvement	−.11	−.08	−.04	.14	.09	−.03	.01	−.06	−.04	−.09	.01	.06	−.06	−.18*	.14†	.23*	-	.32*	.03
18. Cohabitation	−.10	−.20*	.15†	.11	−.07	−.08	.11	−.01	−.27*	−.32*	.23*	.25*	−.33*	−.35*	.20*	.22*	.46*	-	.05
19. Age first sex	−.28*	.18*	.30*	−.03	.14	.11	−.19†	.05	.17†	.06	−.22*	−.21*	.11	.21*	−.09	−.11	−.02	−.13	-
Younger sibling Ma	.51	−.01	19.12	.93	6.81	7.32	4.45	3.99	3.94	3.84	2.71	2.69	2.95	2.48	2.78	2.77	.48	.09	15.80
Younger sibling SD	.50	.83	4.99	.26	1.56	1.21	1.68	1.62	.71	.78	.87	.99	1.09	1.06	.96	1.01	.50	.28	1.19
Older sibling Ma	.50	−.01	19.12	.93	6.81	7.32	4.44	3.99	3.96	3.59	2.61	2.58	2.82	2.40	2.79	2.60	.68	.32	16.58
Older sibling SD	.50	.83	4.99	.26	1.56	1.21	1.68	1.62	.79	.98	.84	.94	1.15	1.00	1.02	1.00	.47	.46	1.78

Note. Younger siblings are above the diagonal and older siblings are below the diagonal. M = mothers. F = fathers. Gender coded as: 0 = females, 1 = males. Parents’ marital status: 0 = not legally married, 1 = legally married. Marital sat = marital satisfaction. Marital con = marital conflict. De-id = de-identification. Rel = Relationship. Relationship involvement, cohabitation: 0 = no, 1 = yes.

^aFor binary variables, proportion for response category = 1 was reported.

^†p < .10.

^*p < .05

Parental Modeling

Starting with the predictors of modeling, examining the constrained versus the unconstrained model for parent gender indicated that the fully constrained model did not fit well, Δχ² (10) = 21.19, p =.02; follow-ups indicated that two paths were not equal across parent gender: parent-youth warmth predicting maternal modeling, Δχ2 (1) = 10.68, p = .00, and parent-youth conflict predicting paternal modeling, Δχ² (1) = 8.05, p = .02. The final model had good fit and explained significant variance in each of the dependent variables (Figure 1). For modeling of mothers, daughters reported higher mean levels than sons, b = −.49, SE = .12, p = .00, and, overall, high levels of maternal warmth but not paternal warmth predicted high levels of maternal modeling five years later. For modeling of fathers, marital satisfaction and parental warmth were positive predictors of modeling, whereas marital conflict was negatively associated with modeling. There were two significant interactions with youth gender by maternal, b = .38, SE = .15, p = .02, and paternal, b = −.29, SE = .14, p = .04, conflict in parent-youth relationship in predicting paternal modeling. Simple slopes analysis revealed that, for sons, high levels of conflict with mother predicted high levels of paternal modeling, but this pattern did not emerge for daughters. The simple slopes analysis for paternal conflict reached trend level for sons, p = .10, but was not significant for daughters, p = .19.

Figure 1.

Significant Unstandardized Coefficients (Standard Errors) from Path Model Linking Marital and Parent-Youth Relationship Qualities (T1) to Youth’s Maternal and Paternal Modeling (T2) and Parents’ Modeling to Romantic Relationships (T2).

Note. Bolded coefficients statistically significant at p < .05. Model was constrained by parent gender and stacked by sibling (N = 492 youth in 246 families). To simplify the figure, control variables (i.e., youth gender, sibling birth order, family SES, parents’ marital relationship duration, parents’ marital status) are not illustrated. AIC = Akaike Information Criteria. ABIC = Sample-Size Adjusted Bayesian Information Criteria.

Moderation findings: Mother and father coefficients were significantly different for parent-youth warmth in predicting maternal modeling (mothers/fathers). Maternal conflict X youth gender interaction predicted paternal modeling (sons/daughters). Paternal conflict X youth gender interaction predicted paternal modeling – but no significant simple slopes; son’s approached significance (italics). Youth’s cohabitation is predicted by a 3-way interaction between parents’ modeling (no difference by parents’ gender) by youth’s gender and birth order (older girls/older boys/younger girls/younger boys). Parental modeling (no difference by parents’ gender) predicted youth’s age of first sex.

Turning to romantic relationship outcomes, specifically, romantic involvement, older siblings, b = .81, SE = .21, p = .00, and females, b = −.63, SE = .25, p = .01, were more likely to be in relationships than were younger siblings and males. Youth’s modeling of parents was unrelated to the likelihood of being in a romantic relationship. Turning to cohabitation, older siblings, b = 1.31, SE = .42, p = .00, and females, b = −1.44, SE = .65, p = .03, were more likely to be cohabitating than younger siblings and males; and high SES was linked to a lower likelihood of cohabitation, b = −.50, SE = .20, p = .01. There was a significant parental modeling by youth gender by birth order interaction effect for cohabitation, b = .48, SE = .18, p = .01. For older siblings and male younger siblings, but not female younger siblings, higher levels of parent modeling were related to lower likelihood of cohabitation. Lastly, for age at first sex, female older siblings, b = −.92, SE = .30, p = .00 (interaction between birth order by gender, b = −.91, SE = .35, p = .01), youth from higher SES families, b = .29, SE = .13, p = .023, and youth in families with longer-married parents, b = .07, SE = .02, p = .00, reported older ages of sexual debut. High levels of youth’s parental modeling also were linked to older age at first sex.

Parental De-identification

The model constrained across parent gender fit as well as the unconstrained model, Δχ² (11) = 6.30, p = .85, suggesting no differences by parent gender. This final model had good fit and explained significant variance (Figure 2). For de-identification from mothers, high levels of marital satisfaction and parent-youth warmth predicted less de-identification five years later. Conversely, high levels of both marital and parental conflict predicted more de-identification five years later. For de-identification from fathers, higher marital satisfaction was associated with lower levels of de-identification. Two interactions also emerged in predicting paternal de-identification. An interaction between maternal-youth conflict by youth gender by birth order, b = .19, SE = .09, p = .04; follow-ups revealed that, for older sons only, more conflict with mothers predicted less de-identification from fathers. A paternal-youth conflict by youth birth order interaction, b = −.23, SE = .08, p = .00 revealed that, for older but not younger siblings, more paternal-youth conflict predicted less de-identification from fathers.

Figure 2.

Significant Unstandardized Coefficients (Standard Errors) from Path Model Linking Marital and Parent-Youth Relationship Qualities (T1) and Youth’s Maternal and Paternal De-identification (T2) and Youth’s De-identification to Youth’s Romantic Relationships (T2).

Note: Bolded coefficients statistically significant at p < .05. Model was constrained by parent gender and stacked by sibling (N = 492 youth in 246 families). To simplify the figure, control variables (i.e., youth gender, sibling birth order, family SES, parents’ marital relationship duration, parents’ marital status) are not illustrated. AIC = Akaike Information Criteria. ABIC = Sample-Size Adjusted Bayesian Information Criteria.

Moderation findings: Paternal de-identification is predicted by a 3-way interaction between mother-youth conflict [difference by parents’ gender: χ²(4) = 89.40, p = .000] by youth’s gender and birth order (older girls/older boys/younger girls/younger boys). Father-youth conflict X youth birth order (older siblings/younger siblings) predicted paternal de-identification. Parental de-identification (no difference by parents’ gender) predicted youth’s romantic involvement. Parental de-identification X youth gender predicted youth’s cohabitation (daughters/sons).

For the romantic relationship outcome variables, starting with romantic involvement, higher levels of parental de-identification were associated with a greater likelihood of being in a romantic relationship. Turning to cohabitation, there was an interaction between parental de-identification by youth gender, b = .13, SE = .06, p = .02. For male, but not female, youth high levels of de-identification were associated with a greater likelihood of cohabitation. Parental de-identification was not a significant predictor of age at first sex.

Discussion

Research and theory have emphasized the role of family experiences in the development of romantic relationships (Bryant & Conger, 2002). This study advanced understanding of potential processes of family influence by measuring youth’s reports of their modeling of and de-identification from their mothers and fathers in the romantic relationship domain, focusing on Mexican-origin youth—an understudied but large and growing subgroup of the U.S. population with early age of first marriage (Payne, 2015). Using an ethnic-homogenous design and multi-informant approach, and consistent with our hypotheses, findings revealed that positivity and conflict in parents’ marriage and in parent-youth relationships predicted youth reports of modeling and de-identifying from their mothers and fathers five years later. Moreover, youth’s reports of parental modeling and de-identification were associated with their concurrent romantic relationship outcomes. Importantly, our results are derived from a sample of Mexican-origin families characterized by long-term couple relationships with moderately high levels of marital satisfaction and low levels of marital conflict—relationships that could set a positive example for youth’s own romantic experiences.

Parental Modeling and De-identification Processes

A contribution of this study was its measurement of parental influence processes: prior work has largely invoked social learning processes of modeling as post hoc explanations for findings of similarities across generations. Guided by theory and research on social learning (Bandura, 1977) and sibling differentiation (Whiteman et al., 2007), and aligning with sociological concepts of modeling and antimodeling (Francis, 1963), we assessed youth’s self-reports of the extent to which they modeled and de-identified from their mothers and fathers in the romantic relationship domain. Importantly, modest negative associations between modeling and de-identification suggested that these were two related but distinct processes: theoretically and empirically, de-identification is not the same phenomenon as low levels of modeling. Our quantitative analyses, revealing that youth reported moderate levels of both modeling and de-identification (i.e., around the midpoint of the scales), corroborate qualitative accounts of these potentially distinct parental influence processes (Lamb & Tamis-Lemonda, 2004). Such findings suggest that prior research focusing only on modeling may underestimate parental influences because youth’s efforts to be different from their parents may cancel out effects of more typically examined social learning mechanisms, yielding only moderate associations between parent and child characteristics. Recognition that youth can learn from their parents’ mistakes (Perez-Brena et al., 2014) or identify parental behaviors that they want to avoid (Francis, 1963) has implications for intervention and prevention efforts aimed at curbing risks and promoting positive youth development, including in the romantic relationship domain.

Predicting Parental Modeling and De-identification Processes

Social learning theory (Bandura, 1977) and perspectives on parental socialization (Darling & Steinberg, 1993) suggest the importance of accounting for parents’ marriages, quality of parent-youth relationships, and gender in youth’s modeling from their parents in the domain of romantic relationships. This study takes a novel step in examining these prospective associations using data from multiple reporters. We found evidence that marriages and parent-youth relationships characterized by more positive qualities predicted higher levels of youth’s self-reported modeling five years later, extending prior work on links between parents’ marital status and youth’s reports of modeling of parents’ relationships (Rhoades et al., 2012). The results for fathers were largely consistent with our predictions, such that higher levels of parental marital satisfaction, lower levels of marital conflict, and higher levels of parental warmth predicted higher levels of youth-reported modeling of fathers. In addition, we found one crossover effect for sons indicating that higher mother-son conflict was associated with more modeling of fathers. For mothers, in contrast, only parental warmth was associated with youth-reported modeling of mothers.

Together, these findings highlight the potentially unique roles of mothers and fathers and the complexity of family dynamics in the Mexican American cultural context. Latino culture is characterized by adherence to traditional gender roles, with fathers serving as providers and authority figures and mothers playing central roles in household and caregiving responsibilities (Cauce & Domenich-Rodríguez, 2002), which may contribute to the overall high levels of maternal relative to paternal modeling. Our findings further suggest that modeling of fathers is strongly situated within the family system and most likely to occur when marital and parent-youth dynamics are more positive. In this sample of parents in long-term relationships, marital quality may be associated with coparenting in the family, which may also spill over to influence father involvement and relationships with children, making fathers accessible and salient models for youth. In the case of youth’s modeling of mothers, who are largely viewed as the primary caretakers within the family (Cauce & Domenich-Rodríguez, 2002), romantic relationship modeling was enhanced by relationship-specific factors (maternal but not paternal warmth) and child characteristics (i.e., youth gender). Given that Mexican culture has a strong emphasis on emotional closeness within the family (Cauce & Domenich-Rodríguez, 2002), warmth in mother-adolescent relationships may be particularly important for the extent to which youth perceive their mothers as models, including for romantic relationships. The salience of maternal warmth may further reflect the more traditional division of parenting roles in this cultural context. It is also notable that youth gender explained modeling of mothers but not fathers. This pattern suggests that maternal influences in romantic relationships, traditionally considered a feminine domain, may be more salient for girls than for boys. Further, because Mexican American fathers are more likely to assume the role of authority figure in the family (Cauce & Domenich-Rodríguez, 2002; Umaña-Taylor & Updegraff, 2012), both female and male youth may look to their fathers as models.

A novel aspect of our study was the examination of predictors of de-identification. Overall, our findings are consistent with our hypotheses, such that youth were more likely to report that they differentiated themselves from both mothers and fathers when parents reported being less satisfied in their marriages; youth whose parents described less parental warmth and more conflict also were more likely to report de-identification from mothers. In other words, when the family emotional climate was less positive, youth reported that they wanted to be more different from their parents in their romantic relationships. Similar to findings on modeling, we found a crossover effect for older sons indicating that higher mother-son conflict was associated with less de-identification from fathers. We did not detect higher levels of de-identification in same- as compared to mixed-gender parent-youth dyads. This is inconsistent with the sibling literature suggesting that youth are more likely to de-identify from individuals who have similar objective characteristics (McHale et al., 2010). More research is needed on parental de-identification processes, which may operate differently than sibling de-identification.

More generally, in contrast to much of the prior literature that highlights links between positive family experiences and positive youth outcomes (i.e., that good things are correlated), our study’s focus on de-identification demonstrates that youth can learn from mistakes and problems in their families. Indeed, de-identification processes may help to explain why some youth exhibit resilience in the face of adverse family circumstances, and our findings suggest that these processes may be an important direction for future study. Our focus here was on de-identification in one specific domain given the significance of forming and maintaining healthy intimate relationships as a developmental task of adolescence and young adulthood (Collins et al., 2009; Masten et al., 2006), but future research should examine whether and how youth perceive that they learn from their parents’ mistakes in other domains of life, such as education, employment, or parenting.

In the case of both modeling and de-identification, findings directed attention to the interconnections among family subsystems (Cox & Paley, 2003) and underscored the utility of simultaneously examining both marital and parent-child relationships and taking into account the perspectives of multiple family members. In general, we found no difference in the overall pattern of associations between marital and parent-adolescent relationship qualities and modeling and de-identification as a function parent gender. The fact that findings were not moderated by parent gender (with two exceptions, noted above) suggests the wide-ranging importance of mothers and fathers. This pattern of findings may also reflect the high levels of interdependence in Mexican-origin families (Cauce & Domenich-Rodríguez, 2002). In addition, our findings for sons illustrate crossover effects, such that relationship experiences as described by one parent predicted youth’s modeling and de-identification with the other parent five years later, suggestive of interdependence among different family subsystems in these influence processes. For sons, higher conflict with mothers was associated with higher modeling of and less de-identification from fathers. The crossover pattern did not emerge for daughters, and this may reflect, in part, the possibility that fathers may not be as involved in the socialization of daughters during this developmental period.

Linking Modeling and De-identification from Parents to Youth’s Relationship Outcomes

Our second goal was to link modeling of and de-identification from parents to youth’s romantic relationship outcomes. We found that higher levels of youth-reported modeling of parents were associated with less likelihood of cohabitation and older age at first sex, but was unrelated to romantic relationship involvement. Further, youth’s self-reported de-identification from parents was associated with higher likelihood of romantic relationship involvement and for sons, a higher likelihood of cohabitation, but unrelated to age at first sex. These results contribute to the literature by providing direct measurement of parental modeling and de-identification and linking youth reports to their concurrent romantic relationship outcomes. Importantly, as noted, we studied these processes in a sample with stable and generally positive couple relationships (as parents reported moderately high levels of marital satisfaction and low marital conflict), which may explain why modeling was associated with more positive outcomes and de-identification was associated with earlier romantic and sexual involvement. In other words, because youth in our sample generally had positive role models, we expected that those who reported modeling would have more positive outcomes. An important next step will be to study modeling and de-identification as linked to romantic relationship outcomes among youth whose parents’ relationships are characterized by relationship challenges such as high conflict or divorce.

More generally, our findings that de-identification was predicted by less warmth and more conflict in parent-adolescent relationships implies that this process reflects a lack of closeness, intimacy, and support inside of the family. Youth may therefore have more need to escape to romantic relationships outside the family of origin (Furman & Shaffer, 2003) and, hence, are more likely to be involved in a romantic relationship and cohabitating with a partner when they report higher levels of de-identification. Because age at first sex may not occur in a relationship context, it may be unrelated to de-identification and youth’s desire to seek supportive relationships outside the family that are distinct from the more conflictual relationships within their family context. As noted, in this sample, youth’s reports of modeling parents, in contrast, appeared to serve as a protective factor, reflecting close and positive relationships within these families. As such, youth may have less inclination to seek support and intimacy from relationships beyond the family, resulting in older ages at first sex and cohabitation. Although modeling was unrelated to youth’s romantic relationship involvement in the regression models, the bivariate association between paternal modeling and youth’s romantic involvement was consistent with the idea that modeling was protective. Understanding the links between modeling and de-identification and romantic relationship outcomes is important as older ages of first sex and cohabitation are beneficial for healthy adjustment (e.g., Kuperberg, 2014), reducing the risks of sexual risk behaviors and related negative health outcomes.

Our finding that de-identification was positively related to cohabitation, but only for sons, may be reflective of parents’ socialization strategies in Latino families. Because of traditional gender norms and attitudes in this cultural context, males are granted more autonomy and choices whereas females tend to spend more time at home, assisting with household and family responsibilities (Cauce & Domenech-Rodríguez, 2002). This is also reflected in our finding that, on average, males reported younger ages at first sex, which is consistent with traditional gender norms that emphasize later sexual debut for girls in this cultural context.

Limitations and Future Directions

In the face of its contributions, the limitations of this study imply directions for future research. First, this study used a sample of Mexican-origin families (two-parent families, mostly from immigrant backgrounds from the Southwest geographic region of the U.S.), and findings may not generalize to other samples of Mexican-origin youth with different family background characteristics or youth from other racial/ethnic backgrounds. It is possible that family experiences have stronger influences on modeling and de-identification dynamics in this cultural group, given the high value placed on family in this cultural context. Replicating findings with other sociocultural groups and testing cross-ethnic similarities and differences in these associations will be important. Second, we studied two-parent families so that we could examine the roles of mothers and fathers, and the parents in our study had been in fairly long-term relationships. Although the majority of Mexican households include two parents (65%; U.S. Census Bureau, 2014), it will be important to extend this work to include other family structures (e.g., divorced, single-parent), as parental influence processes may differ across family structures. As we noted, the positive links between parental modeling and romantic relationship outcomes in this sample may be due to the relationship stability and moderately high marital satisfaction of these parents, and a direction for future research is to study families that differ in parents’ couple relationship experiences.

We measured youth’s reports of modeling and de-identification at one point in time, concurrently with relationship outcomes. This resulted in an informative, but broad, pattern of associations; a formal test of mediation was, however, not ideal because of the temporal ordering of the variables: our measures of modeling and de-identification were collected concurrently with the romantic relationship outcomes, so we were not in the best position to disentangle the direction of the effect. An important next step is to test longitudinal associations between parental modeling and de-identification and youth’s romantic relationship experiences, using additional measures of romantic relationship experiences. Because not all youth in this sample were involved in a romantic relationship, we focused on three romantic relationship outcomes (involvement, cohabitation, age at first sex), allowing us to capture the experiences of all youth in the sample. Future research should include indicators of youth’s romantic relationship quality using both self- and partner-report data.

In addition, although a strength of the study was to study modeling and de-identification in a domain-specific way, our findings may not generalize to other domains of interest. Future work should extend this to other salient tasks of adolescence/young adulthood, such as work and parenting. These kinds of studies will increase our understanding of the intergenerational transmission of family patterns.

Conclusion

The present study adds to the literature on family influences by examining the correlates of Mexican-origin youth’s reports of modeling and de-identification from mothers and fathers in the romantic relationship domain. We focused on an understudied ethnic minority group that is characterized by a strong emphasis on family-oriented values and traditional gender roles (Cauce & Domenech-Rodríguez, 2002) during a critical period of development for establishing and maintaining intimate relationships (Masten et al., 2006). Beyond social learning mechanism of modeling, our findings also reveal the significance of the little studied process of de-identification from parents. Our results highlight the complex family systems, including multiple processes and connections between subsystems in understanding intergenerational transmission processes. Future research should continue the effort to assess parental influence processes and understand how these influence processes operate and are associated with individual, family, and broader sociocultural contextual characteristics.