Introduction
There is compelling evidence that parenting attitudes and behaviors are related to one’s own experiences of being parented in childhood (Bailey et al., Reference Bailey, Hill, Oesterle and Hawkins2009; Belsky et al., Reference Belsky, Jaffee, Sligo, Woodward and Silva2005; Erzinger & Steiger, Reference Erzinger and Steiger2014; Kerr et al., Reference Kerr, Capaldi, Pears and Owen2009; Neppl et al., Reference Neppl, Diggs and Cleveland2020; Savelieva et al., Reference Savelieva, Keltikangas-Järvinen, Pulkki-Råback, Jokela, Lipsanen, Merjonen, Viikari, Raitakari and Hintsanen2017; Thompson et al., Reference Thompson, Jones, Litrownik, English, Kotch, Lewis and Dubowitz2014), a process referred to as the intergenerational transmission of parenting. However, nearly all studies in this area have used retrospective or passive longitudinal designs. To our knowledge, only one study has examined whether intervention-induced improvements in parenting in one generation affect their offspring’s parenting attitudes. In this study, Mahrer et al. (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) found that participation in a parenting-focused program for divorced mothers (G1) when their offsprings (G2) were in late childhood/early adolescence led to an increase in G2’s attitudes toward parental warmth 15 years after the program. There were also interactive effects of G1’s pretest harsh parenting and pretest parental warmth with intervention condition on G2’s. Interactive effects showed that the program especially benefited G2s whose mothers reported either low warmth or high harsh parenting at program entry.
Researchers have identified pathways that may account for the intergenerational transmission of parenting. In their meta-analysis of behavioral genetic influences on parenting, Klahr and Burt (Reference Klahr and Burt2014) demonstrated significant passive and evocative genetic influences on various dimensions of parenting behavior. Consistent with social learning theory (Bandura, Reference Bandura1977), a significant body of research has demonstrated direct transmission of G1 parenting to G2 parenting for harsh parenting (e.g., verbal and physical aggression; Capaldi et al., Reference Capaldi, Pears, Patterson and Owen2003, Reference Capaldi, Pears, Kerr and Owen2008; Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009; Simons et al., Reference Simons, Whitbeck, Conger and Wu1991). A smaller body of research has shown the various dimensions of G1 positive parenting (e.g., involvement, affection, attachment) predict G2 positive parenting (Belsky et al., Reference Belsky, Jaffee, Sligo, Woodward and Silva2005; Kerr et al., Reference Kerr, Capaldi, Pears and Owen2009; Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009). In passive longitudinal studies, results have shown indirect or cascade effects of G1 positive parenting on G2 parenting through G2 functioning, such as peer competence and externalizing problems (Capaldi et al., Reference Capaldi, Pears, Patterson and Owen2003; Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009; Shaffer et al., Reference Shaffer, Burt, Obradović, Herbers and Masten2009). In Mahrer et al.s’ (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) experimental study, the program effects on G2’s attitudes toward warm parenting 15 years after G1 program participation were partially mediated by program-induced effects on G1’s warm parenting. Also, G2’s academic competence and externalizing problems in adolescence were significant mediators of the relation between program effects on G1’s harsh parenting and G2’s attitudes toward warm parenting 15 years later.
Experimental intervention studies, such as Mahrer et al.’s (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) and the current one, offer a robust means of examining the core causal models underpinning resilience theory (Masten & Palmer, Reference Masten, Palmer and Borstein2019; Masten, Reference Masten2007, Reference Masten2014). In resilience theory, parenting is viewed as a key protective resource (e.g., Masten & Palmer, Reference Masten, Palmer and Borstein2019) and a critical leverage point for changing children’s behaviors and attitudes, with cascading effects that spread across time, domains of function, and systems (Masten & Cicchetti, Reference Masten, Cicchetti and Cicchetti2016; Masten, Reference Masten2014; Masten & Palmer, Reference Masten, Palmer and Borstein2019).
In the current study, we used data from multiple assessments over 15 years from G2 childhood/adolescence to emerging/young adulthood that were collected in the randomized controlled trial of the Family Bereavement Program (FBP). The FBP is a preventive intervention for parentally bereaved families that targets parent–child relationship quality, parent demoralization, parent discipline practices, and child coping. In the following sections, we first review the research on the link between attitudes toward parenting, parenting behaviors, and outcomes. Next, we discuss the theoretical basis and empirical support for three plausible mediational cascade pathways that may account for an association between improvements in parental warmth in one generation (G1) and parenting attitudes in the next generation (G2): G1 modeling of parental warmth, G2 functioning (specifically, competencies and externalizing problems), and G2 grief. Then, we discuss the current study.
Relation between parenting attitudes and parenting behaviors and children’s outcomes
Previous research has demonstrated that parenting attitudes and parenting behaviors are strongly linked (Cappa & Dam, Reference Cappa and Dam2014; Kelmendi et al., Reference Kelmendi, Arënliu and Halimi2022; Simons et al., Reference Simons, Beaman, Conger and Chao1993; Vittrup et al., Reference Vittrup, Holden and Buck2006; Xing et al., Reference Xing, Wang, Wang and Wang2019). Parenting attitudes, such as parents’ empathy toward children’s needs or their beliefs about the role of physical punishment, are significantly related to parenting behaviors, such as parental responsiveness, parenting style, neglect, physical punishment, and educational involvement (Bower-Russa, Reference Bower-Russa2005; O’Callaghan et al., Reference O’Callaghan, Borkowski, Whitman, Maxwell and Keogh1999; Oyserman et al., Reference Oyserman, Bybee, Mowbray and MacFarlane2002; Thompson et al., Reference Thompson, Jones, Litrownik, English, Kotch, Lewis and Dubowitz2014). Parenting attitudes are also related to child outcomes such as internalizing and externalizing symptoms, emotion regulation, executive functioning, intelligence, adjustment, and empathy (Babcock Fenerci et al., Reference Babcock Fenerci, Chu and DePrince2016; Kiang et al., Reference Kiang, Moreno and Robinson2004; Miller et al., Reference Miller, Miceli, Whitman and Borkowski1996; Schatz et al., Reference Schatz, Smith, Borkowski, Whitman and Keogh2008; Thompson et al., Reference Thompson, Hollis and Richards2003; Treat et al., Reference Treat, Sheffield Morris, Williamson, Hays-Grudo and Laurin2019).
Plausible cascade pathways
Modeling of parental warmth
It may be that program-induced improvements in G1 parental warmth affect G2’s parenting attitudes through observational learning or modeling. Consistent with social learning theory (Bandura, Reference Bandura1977), children may internalize attitudes regarding how parents should engage with their children (e.g., Capaldi et al., Reference Capaldi, Pears, Patterson and Owen2003, Reference Capaldi, Pears, Kerr and Owen2008) and therefore endorse attitudes about parenting that reflect how they were treated by their parents. Support for this pathway is provided by the findings of Chen and Kaplan’s (Reference Chen and Kaplan2001) three-wave study, which found a significant relation between G1 positive parenting during G2 adolescence and G2 positive parenting practices during their 20s and 30s. This effect remained significant after accounting for three other mediators: G2 psychological state, interpersonal relationships, and social participation. Other longitudinal studies have found that G1’s use of supportive parenting and physical punishment significantly predicts G2’s supportive parenting and physical punishment, respectively (Simons et al., Reference Simons, Beaman, Conger and Chao1993). In one study, G1’s aggressive parenting similarly predicted G2’s aggressive parenting (Conger et al., Reference Conger, Neppl, Kim and Scaramella2003).
G2 functioning (competences and externalizing problems)
From the perspective of a cascade effects model (Masten et al., Reference Masten, Burt, Coatsworth, Cicchetti and Cohen2015; Rutter & Sroufe, Reference Rutter and Sroufe2000; Sameroff, Reference Sameroff2000), it is plausible that the intergenerational transmission of parenting occurs as a result of G1 parenting impacting aspects of G2’s functioning, which in turn affects G2 attitudes toward parenting. Four aspects of G2 functioning have been empirically supported as plausible mediators: social relations (i.e., peer competence and romantic attachment), academic competence, and externalizing problems.
Social relationships: peer competence and romantic attachment. Many studies have linked quality of G1 parenting with G2 peer competence (Eisenberg et al., Reference Eisenberg, Fabes and Murphy1996; Engels et al., Reference Engels, Deković and Meeus2002; Ladd, Reference Ladd2005; Lengua et al., Reference Lengua, Honorado and Bush2007; Lindsey & Mize, Reference Lindsey and Mize2001; Taylor et al., Reference Taylor, Conger, Robins and Widaman2015) and with G2 romantic attachment (Chopik et al., Reference Chopik, Moors and Edelstein2014; Dinero et al., Reference Dinero, Conger, Shaver, Widaman and Larsen-Rife2008; Fraley et al., Reference Fraley, Roisman, Booth-LaForce, Owen and Holland2013; McDowell & Parke, Reference McDowell and Parke2009; Nosko et al., Reference Nosko, Tieu, Lawford and Pratt2011; Zayas et al., Reference Zayas, Mischel, Shoda and Aber2011). In a one-year longitudinal study with a sample of school-aged children, McDowell and Parke (Reference McDowell and Parke2009) found that parent-child interactions characterized by more G1 warmth and G1 positive responses predicted higher G2 teacher- and peer-rated likeability and social competence one year later. In another example, Dinero et al. (Reference Dinero, Conger, Shaver, Widaman and Larsen-Rife2008) found that positive parent-child interactions in G2 adolescence predicted G2 romantic attachment security in emerging adulthood.
There is also evidence that G2 peer competence and romantic attachment are associated with subsequent G2 parenting behaviors. For example, Shaffer et al. (Reference Shaffer, Burt, Obradović, Herbers and Masten2009) found that the transmission of high-quality parenting, which was assessed during G2 young adulthood, was fully mediated by G2 social competence in emerging adulthood. This effect held across gender and ethnicity and remained significant after controlling for G2 IQ and SES. In their 32-year longitudinal study, Raby et al. (Reference Raby, Roisman, Fraley and Simpson2015) found that G1 sensitive caregiving in the first three years of G2’s life predicted G2 peer competence in childhood and adolescence, which in turn predicted G2 romantic relationship competence in young adulthood and supportive parenting in adulthood. A review of more than 60 studies found that insecure romantic attachment was related to providing less sensitive, supportive, and responsive parenting (Jones et al., Reference Jones, Cassidy and Shaver2015).
The relations between quality of G1 parenting, G2 peer and romantic competence, and G2 parenting can be understood from an attachment theory perspective (Bowlby, Reference Bowlby1982). This theory posits that early attachment between children and their parents contributes to the development of an internal working model that is applied to later situations, such as relationships with romantic partners and peers (e.g., Feeney et al., Reference Feeney, Noller, Roberts, Andersen and Guerrero1996). For example, receiving warm parenting may promote the development of skills that support children in having generally positive relationships with others. Further, competence with peers may promote the development of critical relational skills, such as empathy, conflict resolution, and perspective-taking. A recent meta-analysis found that both parent-child and peer relationship quality were related to increased concern for others and increased understanding of others’ emotions (Boele et al., Reference Boele, Van der Graaff, de Wied, Van der Valk, Crocetti and Branje2019), which are important for the positive development and maintenance of all relationships, including parent-child relationships (Collins & Van Dulmen, Reference Collins and Van Dulmen2006; Hartup, Reference Hartup1996; Sroufe, Reference Sroufe2005; Stern et al., Reference Stern, Borelli and Smiley2015). Further, Rostad and Whitaker (Reference Rostad and Whitaker2016) found that a G2 parent’s ability to consider the perspective of their children was a strong predictor of G2 positive parenting (defined as involvement, communication, discipline practices, and support) above and beyond G1 parental rejection, G2 relationship attachment, and G2 mental health problems.
Academic competence. A significant body of literature has established that G1-positive parenting is associated with increased G2 academic performance, engagement, achievement, and attainment (e.g., Brennan et al., Reference Brennan, Shelleby, Shaw, Gardner, Dishion and Wilson2013; Davis-Kean, Reference Davis-Kean2005; Lamborn et al., Reference Lamborn, Mounts, Steinberg and Dornbusch1991; Steinberg et al., Reference Steinberg, Lamborn, Dornbusch and Darling1992). A recent meta-analysis reported that both cross-sectional and longitudinal studies have found that higher G1 parental warmth is associated with better G2 academic performance, whereas G1 parental harsh control is associated with lower G2 academic achievement. Changes in G1 parenting have predicted changes in G2 academic achievement over time (Pinquart, Reference Pinquart2016). In turn, G2 academic achievement has been linked with subsequent G2 parenting behaviors. In a longitudinal study across two generations, Neppl et al. (Reference Neppl, Conger, Scaramella and Ontai2009) found that G2 grade point average in early adolescence significantly mediated the relation between G1’s positive parenting and G2’s positive parenting. Similarly, Mahrer et al. (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) found that G2 high-school grade point average predicted higher G2 warm parenting attitudes in emerging adulthood. Further, research has shown that G2 educational attainment is related to increased G2 parental investment in children (Conger & Donnellan, Reference Conger and Donnellan2007).
Externalizing problems. There is consistent evidence that G1 parental warmth is related to lower G2 externalizing problems. A recent meta-analysis (Pinquart, Reference Pinquart2017) found that higher G1 parental warmth was associated with lower G2 externalizing problems in children both cross-sectionally and longitudinally, and that G1 parental warmth predicted changes in G2 externalizing problems over time. There is also evidence that G2 externalizing problems predict subsequent G2 parenting behaviors. In a longitudinal examination of parenting behaviors over two generations, G2 externalizing behaviors mediated the relation between G1 and G2 harsh parenting (Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009). Similarly, Mahrer et al. (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) found that G2 externalizing problems in adolescence predicted lower G2 warmth attitudes in emerging adulthood.
Further, research has demonstrated that there is a significant genetic basis for the development of G2 outcomes, including social competencies (e.g., Edelbrock et al., Reference Edelbrock, Rende, Plomin and Thompson1995), academic grade point average (e.g., Johnson et al., Reference Johnson, McGue and Iacono2006), and externalizing problems (e.g, Burt, Reference Burt2009). Research has also demonstrated gene-by-environment interactions, in which G2 genetic factors interacted with G1 parenting to predict G2 outcomes in adolescence and adulthood. For example, in an 18-year longitudinal study, genetic polymorphism in the serotonin receptor gene interacted with changes in maternal sensitivity over time to predict romantic attachment in adulthood (Fraley et al., Reference Fraley, Roisman, Booth-LaForce, Owen and Holland2013). In another longitudinal study, parental monitoring during adolescence interacted with polygenic scores to predict externalizing disorders, such that variance in externalizing explained by genetic factors was higher at low levels of parental monitoring (Salvatore et al., Reference Salvatore, Aliev, Bucholz, Agrawal, Hesselbrock, Hesselbrock, Bauer, Kuperman, Schuckit, Kramer, Edenberg, Foroud and Dick2015).
Grief
For bereaved children, there is theoretical and empirical support for a cascade effects model in which intervention-induced improvements in G1 parenting lead to reductions in G2 grief and, in turn, affect G2 attitudes toward parenting in young adulthood. Alvis et al. (Reference Alvis, Dodd, Oosterhoff, Hill, Rolon-Arroyo, Logsdon, Layne and Kaplow2022a) proposed theoretical processes through which parenting might affect their offspring’s grief, though few studies have examined this relation empirically. They proposed that G1 parental responsiveness and sensitivity to their G2 children’s negative affect may help G2s to cope with their distress over the death of their parent or the secondary stressors that follow the death. For example, responsive G1 parents might provide emotion coaching by validating and empathizing with their child’s feelings rather than minimizing and invalidating their children’s negative emotions (Katz et al., Reference Katz, Maliken and Stettler2012). We are aware of only three studies that have empirically assessed the relations between aspects of parenting and children’s grief. Shapiro et al. (Reference Shapiro, Howell and Kaplow2014) assessed caregiver behaviors during a discussion with their bereaved children concerning positive memories of their deceased parent. They found that G1-positive parenting behaviors (defined as sensitivity to children’s needs, positive engagement, warmth, positivity, ease of conversation, and conversational depth), were related to lower symptoms of G2’s maladaptive grief. Alvis et al. (Reference Alvis, Zhang, Sandler and Kaplow2022b) found that G2 reports of their G1 caregiver’s avoidance and inhibition of grief discussions were related to higher levels of G2 maladaptive grief. In the only prospective longitudinal study with bereaved children, Wolchik et al. (Reference Wolchik, Ma, Tein, Sandler and Ayers2008) found that caregiver–child relationship quality was significantly related to children’s lower intrusive grief thoughts 11 months later.
Although there are no empirical studies on the impact of grief experienced during childhood/adolescence on attitudes toward parenting or parenting behaviors in adulthood, we hypothesized that increased grief could lead to more maladaptive parenting attitudes because of relations between childhood grief and adult outcomes that are cross-sectionally associated with parenting. For example, a study with parentally bereaved children/adolescents found that childhood grief had an indirect effect on worsened major depression (Sandler et al., 2023) and a direct effect on higher levels of suicidal ideation/attempts (Sandler et al., Reference Sandler, Tein, Zhang, Wolchik and Thieleman2021) 14 years later during adulthood. Given that depression and suicidality have been shown to relate cross-sectionally to less positive and more negative parenting behaviors in multiple studies (e.g., Lovejoy et al., Reference Lovejoy, Graczyk, O’Hare and Neuman2000), it follows that childhood grief may have an effect on parenting attitudes during adulthood.
Current study
In this study, we used data from a randomized controlled trial of the Family Bereavement Program (FBP), a preventive intervention for bereaved families, to examine the program’s effect on G2 parenting attitudes 15 years after the program. We tested the direct effects of the FBP on G2 parenting attitudes at the 15-year follow-up and cascade effects through potential mediating pathways at intervening assessment periods. For the theoretical model, see Figure 1. Each cascade effect model included three paths: (1) effects from intervention to posttest (T2) G1 parental warmth; (2) effects from T2 G1 parental warmth to the potential mediator at the six-year follow-up (T4); and (3) effects from the T4 potential mediator to the 15-year follow-up (T5) G2 attitudes toward warm parenting and attitudes toward physical punishment. Potential mediators at T4 included G1 parental warmth, G2 functioning (i.e., academic competence, peer competence, romantic attachment, and externalizing problems), and G2 grief (i.e., posttraumatic growth through grief and grief-related social detachment/insecurity). Examining whether a preventive parenting intervention affects the parenting attitudes of young adults who have just started to have families or do not yet have children is important, given that parenting attitudes are established before or early in parenting (Powell & Karraker, Reference Powell and Karraker2017) and are significant predictors of subsequent parenting behavior (Cappa & Dam, Reference Cappa and Dam2014; Kelmendi et al., Reference Kelmendi, Arënliu and Halimi2022; Simons et al., Reference Simons, Beaman, Conger and Chao1993; Vittrup et al., Reference Vittrup, Holden and Buck2006; Xing et al., Reference Xing, Wang, Wang and Wang2019).
Figure 1. The hypothesized model of how the FBP may increase G2 Warmth attitudes and decrease G2 Physical Punishment attitudes through direct, modeling, and cascading effects.
Note. *Adolescent functioning refers includes improvements in academic competence, peer competence, anxious and avoidant romantic attachment, and externalizing problems.
The experimental nature of the study and its multiple assessments over 15 years are important features of this study. Although multiple longitudinal studies have demonstrated the continuity of parenting behaviors across generations, a randomized experimental design provides a more rigorous examination of the associations between G1 parenting and G2 attitudes toward parenting and allows testing of the role that parenting plays in resilience theory (Masten, Reference Masten2007, Reference Masten2014). Randomization disentangles the effects of intervention-induced changes in G1 parenting from variables in the environment that impact this relation in non-experimental studies (e.g., parent depression, shared genes). In addition, experimental studies address the generational shift in which recent generations of parents have increasingly shifted toward using more parental warmth and using less strictness (including both harsh and nonphysical punishment practices) in their parenting practices (Garcia et al., Reference Garcia, Fuentes, Gracia, Serra and Garcia2020).
As reported in our pre-registered plan, we hypothesized the following:
H1: Direct effects
There will be a direct effect of participation in the FBP on G2’s parenting attitudes 15 years later such that G1 participation in the FBP would lead to G2s having higher warmth attitudes and lower physical punishment attitudes. Further, program effects on G2 parenting attitudes will be stronger for G2s whose parents entered the program with less parental warmth.
H2: Cascade effects through G1 parental warmth
Intervention-induced increases in G1 parental warmth at posttest will lead to stability or improvements in G1 parental warmth at the 6-year follow-up which will, in turn, lead to higher G2 warmth attitudes and lower physical punishment attitudes at the 15-year follow-up.
H3: Cascade effects through G2 functioning
Intervention-induced increases in G1 parental warmth at the posttest will lead to improvements in G2’s romantic attachment, peer competence, and academic competence, and to decreased externalizing problems at the 6-year follow-up, which will lead to higher G2 warmth attitudes and lower physical punishment attitudes at the 15-year follow-up.
H4: Cascade effects through G2 grief
Intervention-induced increases in G1 parental warmth at the posttest will lead to improvements in G2 growth through increased posttraumatic grief and decreased grief-related social detachment/insecurity at the 6-year follow-up, which will lead to G2 higher warmth attitudes and lower physical punishment attitudes at the 15-year follow-up.
Exploratory analyses
In exploratory analyses, we tested whether these models differed by G2 gender (male vs. female) or age (8–12 vs. 13–16). We examined whether G2 gender moderated the direct and cascade effects based on prior evidence of G2 gender differences in the transmission of parenting attitudes and parenting behaviors, although results are inconsistent regarding whether transmission is stronger for females (Belsky et al., Reference Belsky, Jaffee, Sligo, Woodward and Silva2005; Isley et al., Reference Isley, O’Neil, Clatfelter and Parke1999; Simons et al., Reference Simons, Beaman, Conger and Chao1992) or males (Madden et al., Reference Madden, Domoney, Aumayer, Sethna, Iles, Hubbard, Giannakakis, Psychogiou and Ramchandani2015; Savelieva et al., Reference Savelieva, Keltikangas-Järvinen, Pulkki-Råback, Jokela, Lipsanen, Merjonen, Viikari, Raitakari and Hintsanen2017; Simons et al., Reference Simons, Beaman, Conger and Chao1992). Given the inconsistencies in the literature, we did not make a priori hypotheses about the direction of these effects. We also examined whether age moderated the direct and cascade effects, given that G2s participated in the program at different developmental stages. Although differential effects of the potential mediators that we examined have rarely been studied by age (for an exception see Tammilehto et al., Reference Tammilehto, Punamäki, Flykt, Vänskä, Heikkilä, Lipsanen, Poikkeus, Tiitinen and Lindblom2021), we thought it was plausible that the impact of the putative mediators of parental warmth, romantic attachment, and peer competence might vary by age, such that the relations between parental warmth and attitudes would be stronger for the younger than older G2s and the relations between peer and romantic competence and attitudes would have stronger for older than younger G2s (Allen et al., Reference Allen, Grande, Tan and Loeb2018). For academic competence, externalizing problems, and grief, we did not have data or theory to support differential effects on parenting attitudes by G2 age.
Method
Participants
Bereaved families were recruited from community agencies, including schools and service agencies, as well as by mail solicitation (for full data collection procedures, please see Sandler et al., Reference Sandler, Ayers, Wolchik, Tein, Kwok, Haine, Twohey-Jacobs, Suter, Lin, Padgett-Jones, Weyer, Cole, Kriege and Griffin2003). There were several eligibility criteria, including: (1) family experienced parental death between four and 30 months before beginning the study (2) family had one or more children between the ages of 8 and 16; (3) family was not currently receiving other mental health or bereavement services; (4) family was willing to participate in either the intervention (FBP) or the literature control self-study (LC) program; (5) parents (used to describe the child’s primary post-bereavement caregiver) and youth were able to complete the assessment in English; (6) youth were not receiving special education services; and (7) family planned to stay in the area for the next six months. Families were referred to mental health treatment services if either the child or parent expressed suicidal intent or if the parent was diagnosed with major depression using the Structured Clinical Interview for DSM–IV (First et al., Reference First, Gibbon, Spitzer and Williams1996). In addition, because of potential problems complying with group procedures, children were excluded and referred for clinical services if they were diagnosed with conduct disorder, oppositional defiant disorder, or attention-deficit/hyperactive disorder (that was not being treated with medication) using the Diagnostic Interview Schedule for Children–Child/Parent Informant (Shaffer et al., Reference Shaffer, Fisher, Dulcan, Davies, Piacentini, Schwab-stone, Lahey, Bourdon, Jensen, Bird, Canino and Regier1996). Following the pretest interview, families were randomized to the FBP (n = 90 families; 135 children) or self-study literature control (LC) condition (n = 66 families; 109 children) at a 55/45 ratio.
The sample was comprised of 244 children and adolescents (G2). The mean age of the children at program entry was 11.4 years (SD = 2.43) and 26.71 (SD = 2.35) at the 15-year follow-up. Fifty-four percent were males. Ethnicity was as follows: 67% non-Hispanic White, 16% Hispanic, 7% African American, 3% Native American, 1% Asian or Pacific Islander, and 6% other. At the pretest, 63% resided with their mothers, 21% with their fathers, and 16% with a nonparental family member or friend. Parental death occurred an average of 10.81 months before the study (SD = 6.35). Sixty-three per cent of the caregivers were mothers, 21% were fathers, and 16% were another relative or friend. Cause of death was 67% illness, 20% accident, and 13% homicide or suicide. Median family income was between $30,000 and $35,000 and 15.9% of families were below the poverty line according to the U.S. Health and Human Services poverty guidelines for 1996 (Sandler et al., Reference Sandler, Ayers, Wolchik, Tein, Kwok, Haine, Twohey-Jacobs, Suter, Lin, Padgett-Jones, Weyer, Cole, Kriege and Griffin2003). Using data from death certificates for adults aged 28–58 years in the county in which the study occurred, Sandler et al. (Reference Sandler, Ayers, Wolchik, Tein, Kwok, Haine, Twohey-Jacobs, Suter, Lin, Padgett-Jones, Weyer, Cole, Kriege and Griffin2003) showed no differences between this sample and the population of deaths in the county in this age range on ethnicity: χ 2(5, N = 148) = 1.47, ns, gender: χ 2(1, N = 153) = 0.17, ns, or cause of death: χ 2(2, N = 153) = 1.28, ns. Prior evaluations have shown that families in the FBP and LC groups were comparable in terms of the demographic variables and pretest variables (Sandler et al., Reference Sandler, Ayers, Wolchik, Tein, Kwok, Haine, Twohey-Jacobs, Suter, Lin, Padgett-Jones, Weyer, Cole, Kriege and Griffin2003), except that the percentage of non-Hispanic Whites was lower in the FBP group than in the LC group (64 vs. 72%).
Intervention conditions
The Family Bereavement Program (FBP) is a 12-session program that includes separate groups for parents, adolescents, and children, plus two individual family sessions. The manualized program was delivered by two master’s level counselors. Fidelity to the program was very high (Sandler et al., Reference Sandler, Ayers, Wolchik, Tein, Kwok, Haine, Twohey-Jacobs, Suter, Lin, Padgett-Jones, Weyer, Cole, Kriege and Griffin2003) such that objective raters reported that over 80% of the action items described in the manual were delivered by group leaders. The parent component focused on teaching skills and activities to promote positive parent–child relationships (e.g., catch ‘em doing good, active listening), supporting parents’ adaptive grief processes (e.g., identifying and progressing toward bereavement-related goals), strengthening effective discipline practices (e.g., clear expectations, consistent and appropriate consequences), and reducing parents own depression and grief (e.g., normalizing grief experiences, increasing involvement in positive activities). The child and adolescent components of the program focused on activities to strengthen effective coping skills (e.g., cognitive reframing, problem-solving), improve the parent-child relationship, adaptively express grief-related feelings, and reduce threat appraisals. For a full description of the FBP, please refer to Ayers et al. (Reference Ayers, Wolchik, Sandler, Twohey, Weyer, Padgett-Jones, Weiss, Cole and Kriege2014) and Sandler et al. (Reference Sandler, Wolchik, Ayers, Tein and Luecken2013).
The LC condition consisted of three developmentally appropriate books on grief sent to the parents, adolescents, and children. Forty-two percent of parents, 38% of adolescents, and 71% of children reported reading 50% or more of the books.
Procedures
Data were collected at five-time points (pretest (T1), posttest (T2), 11-month follow-up (T3), 6-year follow-up (T4), and 15-year follow-up (T5). This study uses data (T1), (T2), (T4), and (T5). Retention rates for the FBP and LC were 98 and 95% (T2); 87 and 94% (T4); and 80 and 73% (T5), respectively. There were no differences in retention rates between the FBP and LC at posttest, 6-, or 15-year follow-ups. After controlling for alpha inflation, attrition analyses at T5 indicated no statistically significant attrition or attrition × group effects related to demographic factors, including parent/child gender, race/ethnicity, parent education, and baseline income, or baseline mental health problems including child internalizing and externalizing problems.
Interviews were conducted in the families’ homes; parents and children were interviewed separately by trained interviewers. After confidentiality was explained, parents and offspring 18 or older provided informed consent, and children provided informed assent. At T1 and T2, families were paid $40 for interviews involving one child and an additional $30 for each additional child who participated in data collection. At T4 and T5, G1 and G2 each received $175.
Interviewers were masked to participants’ group assignment at all time points, and participants were instructed not to disclose their group assignment to interviewers. At T4, interviewers’ knowledge of condition was assessed; 96.5% of interviewers reported that they did not know the interviewee’s program condition. All procedures were approved by the university’s Institutional Review Board.
Measures
Mediators
G1 parental warmth. G1 parental warmth was assessed at T1, T2, and T4 using four measures. Both G1 and G2 completed the two subscales of the Child Report of Parental Behavior Inventory (CRPBI; Schaefer, Reference Schaefer1965): the 16-item Acceptance subscale (e.g., “Your parent enjoyed doing things with you”, α = .90–.93; range across the assessments are reported for all alphas) and 16-item Rejection subscale (e.g., “Your parent said you were a big problem”, α = .81–.90). Both G1 and G2 reports of the CRPBI have demonstrated adequate reliability and validity (e.g., Schaefer, Reference Schaefer1965; Wolchik et al., Reference Wolchik, West, Sandler, Tein, Coatsworth, Lengua, Weiss, Anderson, Greene and Griffin2000). In addition, G1 completed the 7-item Talk with Reassurance subscale of the Caregiver Expression of Emotion Questionnaire (Jones & Twohey, Reference Jones and Twohey1998; e.g., “Reassure child that you are dealing with your sadness”; α = .74–.85) and G2 completed the 10-item Sharing of Feelings scale (Ayers et al., Reference Ayers, Sandler and Twohey1998; e.g., “Your parent understands your feelings”; α = .83–.91). Prior measurement work with confirmatory factor analysis showed that the one-dimensional model of parental warmth adequately fit the data: T1: χ2(4, N = 204) = 9.44; CFI = .98, RMSEA = .08; SRMR = .04; T2: χ2(4, N = 197) = 6.07; CFI = .99, RMSEA = .06; SRMR = .03; T4: χ2(4, N = 178) = 4.33; CFI = .99, RMSEA = .02; SRMR = .01 (see Sandler et al., Reference Sandler, Tein, Cham, Wolchik and Ayers2016a). Thus, a composite of these measures was used as the score for parental warmth.
G2 peer competence. G2 peer competence was measured using G2 report at T4 on the 7-item Peer Relationships subscale of the Coatsworth Competence Scale (Coatsworth & Sandler, Reference Coatsworth and Sandler1993; e.g. “You are liked by lots of peers your age”; α = .63–.78). Composite scores across G1 and G2 reports were created for the pretest measures by computing the mean of z-scores across G1 and G2 reports. The Coatsworth Competence Scale has been shown to have adequate convergent and discriminant validity and internal consistency reliability in community and at-risk samples (Coatsworth & Sandler, Reference Coatsworth and Sandler1993; Spaccarelli et al., Reference Spaccarelli, Coatsworth and Bowden1995).
G2 academic competence. G2 academic competence was assessed using G1 and G2 reports at T1 and G2 report at T4 on the Coatsworth Competence Scale 6-item Academic Competence subscale (Coatsworth & Sandler, Reference Coatsworth and Sandler1993; e.g. “You got mostly A’s and B’s in school”; α = .81–.89). Composite scores across G1 and G2 reports were created for the T1 measures by computing the mean of z-scores across G1 and G2 reports. The Coatsworth Competence scale has been shown to have adequate convergent and discriminant validity and internal consistency reliability in community and at-risk samples (Coatsworth & Sandler, Reference Coatsworth and Sandler1993; Spaccarelli et al., Reference Spaccarelli, Coatsworth and Bowden1995).
G2 romantic attachment. At T4, G2 completed two subscales of the Experiences in Close Relationships Scale (ECR; Brennan et al., Reference Brennan, Clark, Shaver, Simpson and Rholes1998) that assess general romantic experiences: the 18-item Anxiety subscale (e.g., “I need a lot of reassurance that I am loved by my partner,” α = .93) and 18-item Avoidance subscale (e.g. “I prefer not to show a partner how I feel deep down,” α = .90). The scale is reliable and valid (Sibley et al., Reference Sibley, Fischer and Liu2005).
G2 externalizing problems. G2 externalizing problems were measured at T1 and T4 using a composite of G1 and G2 reports on the Child Behavior Checklist (CBCL; Achenbach & Edelbrock, Reference Achenbach and Edelbrock1991) and Youth Self-report (YSR; Achenbach, Reference Achenbach1991) respectively for G2s younger than 18 years. For G2s aged 18 and older, G1s completed the Young Adult Behavior Checklist (YABCL; Achenbach, Reference Achenbach1993) and G2s completed the Young Adult Self-report (YASR; Achenbach, Reference Achenbach1990). Because the measures for G2 adolescents and young adults are not identical, previous measurement work performed by this team (see Reference Sandler, Tein, Cham, Wolchik and Ayers2016a, Sandler et al., Reference Sandler, Ma, Tein, Ayers, Wolchik, Kennedy and Millsap2010) applied item response theory to conduct an equating transformation that selected conceptually equivalent items and put the scale scores (across CBCL and YABCL and across YSR and YASR, respectively) on a common metric using a large data set obtained from Achenbach (Thomas M. Achenbach, Ph.D., unpublished raw data from the CBCL, YABCL, YSR, and YASR, 2003) that contained self- and parent-report scores on the CBCL/YABCL and YSR/YASR. The resulting 35-item CBCL, 34-item YABCL, 32-item YSR, and 27-item YASR subscales had good reliability, with T4 internal consistencies of .92, .93, .88, and .87, respectively. At T1, CBCL, and YSR internal consistencies were .90 and .86, respectively. The CBCL and YABCL have good reliability and validity (Achenbach & Edelbrock, Reference Achenbach and Edelbrock1991; Achenbach, Reference Achenbach1993).
G2 posttraumatic growth through grief. At T4, G2s completed the 7-item Relating to Others subscale of the Posttraumatic Growth Inventory (PTGI; Tedeschi & Calhoun, Reference Tedeschi and Calhoun1996; e.g. “You have a greater sense of closeness with others”; α = .91). This scale assessed how G2s felt that their relationships with others had improved as a result of dealing with their parent’s death. Tedeschi and Calhoun (Reference Tedeschi and Calhoun1996) reported adequate test-retest reliability and internal consistency of this measure as well as concurrent and discriminant validity. In this sample, the Relating to Others subscale was significantly inversely correlated with measures of relationship avoidance (r = −.38) and grief-related Social Detachment/Insecurity (r = −.43).
G2 grief-related social detachment/insecurity. At T4, G2s completed the Social Detachment/Insecurity Subscale (Kennedy, Reference Kennedy2006; α = .83), which is a composite of seven items from the 26-item Inventory of Complicated Grief (ICG; Prigerson et al., Reference Prigerson, Maciejewski, Reynolds, Bierhals, Newsom, Fasiczka, Frank, Doman and Miller1995) that assess feelings of loneliness, a lost sense of safety, trust and control, numbness, jumpiness, and perceived distance from others (e.g., “Over the past month, to what extent has it been hard for you to trust others?”). Social Detachment/Insecurity is significantly associated concurrently with G2 and G1 reports of internalizing and externalizing problems and inversely related to self-esteem and peer competence.
Outcomes
G2 parenting attitudes. G2 attitudes toward parental warmth were measured at the 15-year follow-up using the 8-item Empathetic Awareness subscale of the Adult-Adolescent Parenting Inventory (Bavolek, Reference Bavolek1985; e.g. “Children will quit crying faster if they are ignored”, α = .88). Attitudes toward physical punishment were measured at the 15-year follow-up using the 10-item Physical Punishment subscale of the Adult-Adolescent Parenting Inventory (Bavolek, Reference Bavolek1985; e.g., “Children deserve more discipline than they get,” α = .92). Both subscales have demonstrated adequate reliability and validity (Bavolek, Reference Bavolek1985; Conners et al., Reference Conners, Whiteside-Mansell, Deere, Ledet and Edwards2006).
Covariates
In all models, we included matched pretest controls of mediator variables (e.g., T1 externalizing was a covariate for models that included T3 externalizing) with two exceptions. First, because it was not developmentally appropriate to ask children about romantic attachment at T1, peer competence, which we determined was the most closely related variable assessed at pretest, was used as a pretest covariate in the models involving romantic attachment. At T1, peer competence was a composite of G1 and G2 reports on the 7-item Peer Relationships subscale of the Coatsworth Competence Scale (Coatsworth & Sandler, Reference Coatsworth and Sandler1993). Second, neither measure of grief was assessed at T1. As such, a composite score (i.e., equally weighted sum of standardized scores) of the 9-item Intrusive Grief Thoughts Scale (IGTS, Program for Prevention Research, 1999, α = .89), which assessed the degree to which current negative grief-related thoughts intruded on children’s everyday lives, and the 13-item Present Feelings subscale of the Texas Revised Inventory of Grief (TRIG; Faschingbauer, Reference Faschingbauer1981; α = .89), which assessed the frequency of intrusive, negative, and disruptive experiences related to grief, was used as a covariate for the two grief outcomes. Reliability and validity of the TRIG are acceptable (Futterman et al., Reference Futterman, Holland, Brown, Thompson and Gallagher-Thompson2010; Holm et al., Reference Holm, Alvariza, Fürst, Öhlen and Årestedt2018; Montano et al., Reference Montano, Lewey, O’Toole and Graves2016). Due to high skewness and kurtosis, two items were dropped at all time points (“sometimes I very much miss my [deceased parent]”; “no one will ever take the place of my [deceased parent] who died”).
T1 parental warmth was included as a covariate as the matched control for T2 parental warmth. We also identified other variables that were significantly related to the intergenerational transmission of parenting behaviors and attitudes in the literature (e.g., Hoff et al., Reference Hoff, Laursen and Tardif2002; Simons et al., Reference Simons, Beaman, Conger and Chao1993) and considered them for inclusion as covariates: T1 income, G2 gender, G2 age, G1 gender, T1 G1 education, composite G1 and G2 reports of T1 G2 internalizing problems, composite G1 and G2 reports of T1 G2 externalizing problems, and G2 having children at the 15-year follow-up (yes-no).
Data analytic strategy
Our data analytic strategy and hypotheses were pre-registered through the Open Science Framework (https://doi.org/10.17605/OSF.IO/NFJST).
We probed for potentially influential cases by looking for cases with a Cook’s distance exceeding .20 (Bollen & Jackman, Reference Bollen and Jackman1985). To determine what covariates to include in the models, we examined correlations between the T4 mediators and T5 dependent variables and the potential covariates. Any covariate that was significantly related to warmth attitudes or physical punishment attitudes was included as a control variable for all models; those that were significantly related to a T4 mediator were included only in models that contained that mediator.
We used the structural equation modeling framework to test the hypothesized models, using Mplus (version 8.3; Muthén & Muthén, Reference Muthén and Muthén1998-2017). The Mplus command TYPE = COMPLEX (i.e., sandwich estimator) was used to adjust standard errors due to the clustering effects of children being nested within families. In all models, variables at the same time point were allowed to covary and we controlled for the matched pretest mediator variables. Given that Mahrer et al. (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) found an interactive effect of intervention with pretest parental warmth on warmth attitudes 15 years later, before running models, we examined whether this interaction predicted G2 warmth attitudes at T5. Per our pre-registration, if the interaction term was a significant predictor of warmth attitudes, it would be included in all mediation models.
For the cascade mediation models, as shown in the theoretical model presented in Figure 1, the proximal mediator (G1 parental warmth) was assessed at T2; the distal mediators (G1 parental warmth and G2 academic competence, peer competence, romantic attachment, externalizing problems, posttraumatic growth through grief, and grief-related social alienation) were assessed at T4; and the dependent variables (G2 warmth attitudes and physical punishment attitudes) were assessed at T5. T1 parental warmth, matched control variables (i.e., pretest measure of the mediator/outcome or proxy), and demographic variables that were significantly related to the mediator or outcomes were included as covariates.
Each hypothesis was tested in a separate path model. Then, to investigate the overall effects of the proximal G1 mediator and the G1 and G2 distal mediators and the unique effect of each mediator over and above the others, we evaluated a model that included the variables that had a significant (p < .05) or marginal (p < .10) path from both T2 parental warmth to the T4 mediator and the T4 mediator to warmth attitudes or attitudes toward physical punishment.
Finally, we conducted two exploratory analyses to assess moderation by G2 age and sex, respectively. We examined a multigroup model in which each hypothesis was explored separately for (1) younger (8–12 years old) vs. older (13–16 years old) G2s at program entry and (2) male vs. female G2s. Given the large number of paths and the lack of a priori hypotheses for the predicted effects, we used the false discovery rate procedure (Benjamini & Hochberg, Reference Benjamini and Hochberg1995) to adjust for multiple tests.
Results
Preliminary analysis
No influential cases were identified. Descriptive statistics and zero-order Pearson product-moment correlations are presented in Tables 1 and 2. G2 age (at the time of the program) and gender, as well as whether G2s had children of their own (yes or no), were significantly correlated with parenting attitudes and thus were included as covariates for all models. G1 gender was significantly related to romantic attachment and thus was included as a covariate in the model that contained romantic attachment. The intervention by pretest parental warmth interaction did not significantly predict warmth attitudes at 15 years, so it was not included in any model.
Table 1. Correlations and descriptive statistics of attitudes and demographic variables
Note. *p ≤ .05; **p ≤ .01. Group was coded as 0 = LC and 1 = FBP. Income was measured in $5,000 categories from 1 to 21 ranging from 1 = <$5,000 to 21 = > $100,00. Gender was coded as 0 = Male and 1 = Female. Education was assessed on a scale of 1–7 ranging from elementary school to a graduate degree. Parent status was coded as 0 = Does not have any children and 1 = Has children.
Table 2. Correlations and descriptive statistics of outcome and potential mediating variables
Note. All correlations greater than or equal to ±.14 are significant at the p ≤ .05 level; all correlations greater than or equal to ±.19 are significant at the p ≤ .01 level. Group was coded as 0 = LC and 1 = FBP.
Warmth attitudes at T5 were positively correlated with T4 academic competence (r = .273, p = .003) and negatively correlated with T4 anxious romantic attachment (r = −.186, p = .024) and externalizing problems (r = −.162 p = .041). Physical punishment attitudes at T5 were positively correlated with T4 externalizing problems (r = .212, p = .007). Physical punishment attitudes were negatively correlated with warmth attitudes (r = −.43, p < .001).
Direct effects of the FBP
The intervention was a significant predictor of G2 physical punishment attitudes such that participation in the FBP predicted decreased physical punishment attitudes (B = −0.149, SE = 0.067, t = −2.214, p = 0.027). There was no direct effect of the intervention on G2 warmth attitudes (B = 0.034, SE = 0.080, t = 0.423, p = .674).
Cascade effects of the FBP
Cascade effects through G1 parental warmth
Participation in the FBP was significantly related to T2 G1 parental warmth (B = 0.140, SE = 0.047; t = 2.999; p = .003) such that those who participated in the FBP reported higher G1 parental warmth. In turn, higher T2 G1 parental warmth predicted higher T4 G1 parental warmth (B = 0.393, SE = 0.083; t = 4.734; p < .001). However, T4 G1 parental warmth was not significantly related to G2 warmth attitudes (B = −0.053, SE = 0.106; t = −0.500; p = .617). It was marginally, positively related to G2 physical punishment attitudes (B = 0.148, SE = 0.090; t = 1.655; p = .098), but in the unexpected direction, such that higher T2 G1 parental warmth was related to marginally higher T5 G2 physical punishment attitudes. The FBP intervention had a direct negative effect on T5 G2 physical punishment (B = −0.174, SE = 0.065, t = −2.656, p = 0.008) after accounting for the mediation effect.
Cascade effects through G2 functioning
Each mediator was tested in an individual model. As noted above, participation in the FBP was significantly related to T2 G1 parental warmth (B = 0.140, SE = 0.046, t = 3.038, p = .003). In turn, T2 G1 parental warmth significantly predicted lower T4 G2 externalizing problems (B = −0.145, SE = 0.065, t = −2.229, p = .026). Higher externalizing problems marginally predicted lower T5 G2 warmth attitudes (B = −0.152, SE = 0.086, t = −1.765, p = .078) but did not significantly predict T5 G2 physical punishment attitudes (B = 0.134, SE = 0.099, t = 1.360, p = .174). Higher T2 parental warmth predicted lower T4 anxious romantic attachment (B = −0.170, SE = 0.078, t = −2.188, p = .029). Higher T4 G2 anxious romantic attachment significantly predicted higher T5 G2 warmth attitudes (B = −0.241, SE = 0.086, t = −2.795, p = .005) and was marginally related to T5 G2 lower physical punishment attitudes (B= 0.140, SE = 0.077, t = 1.813, p = .070). T2 parental warmth did not significantly predict T4 G2 academic competence (B = 0.112, SE = 0.089, t = 1.253 p = .210), peer competence (B = −0.005, SE = 0.071, t = −0.068, p = .946), or avoidant romantic attachment (B = −0.062, SE = 0.067, t = −0.932, p = .351). Higher T4 academic competence predicted higher T5 G2 warmth attitudes (B = 0.224, SE = 0.085, t = 2.636, p = .008) but not physical punishment attitudes (B = −0.142, SE = 0.094, t = −1.514, p = .130). T2 parental warmth did not predict T4 G2 peer competence or avoidant romantic attachment nor did T4 G2 peer competence or avoidant romantic attachment significantly predict T5 G2 attitudes toward warmth or physical punishment. In all the above models, the direct effect of the FBP on decreased physical punishment attitudes remained significant.
Cascade effects through G2 posttraumatic growth through grief and grief-related social detachment
Participation in the FBP was significantly related to T2 G1 parental warmth (B = 0.14, SE = 0.046; t = 3.038; p = .003). T2 parental warmth did not significantly predict T4 G2 posttraumatic growth through grief or grief-related social detachment. Further, neither T4 G2 posttraumatic growth through grief nor grief-related social detachment predicted T5 G2 warmth or physical punishment attitudes.
Model combining warmth, anxious romantic attachment, and externalizing problems
We evaluated a combined mediator model that included the variables that had a significant (p < .05) or marginal (p < .10) path from both T2 G1 parental warmth to the T4 G2 mediator and from the T4 G2 mediator to at least one of the T5 G2 attitudes variables. Of all mediators tested, parental warmth, anxious romantic attachment, and externalizing problems met these conditions (see Figure 2).
Figure 2. Cascade effects of G1 parental warmth, G2 romantic attachment, and G2 externalizing problems on G2 attitudes toward parenting in emerging/young adulthood.
Note. *p ≤ .05; **p ≤ .01; † p ≤ .10. Intervention condition was coded as 0 = LC and 1 = FBP.
In this model, participation in the FBP remained directly and significantly related to lower T5 G2 physical punishment attitudes (B = −0.13, SE = 0.063, t = −2.048, p = .041) above and beyond the mediator effects. Consistent with previous findings, participation in the FBP was significantly related to T2 G1 parental warmth (B = 0.14, SE = 0.047, t = 2.998, p = .003). In turn, higher T2 G1 warmth predicted lower T4 G2 anxious romantic attachment (B = −0.172, SE = 0.078, t = −2.220, p = .026) and lower T4 G2 externalizing problems (B = −0.136, SE = 0.066, t = −2.064, p = .039). Higher T4 G2 anxious romantic attachment predicted T5 G2 lower warmth attitudes (B = −0.217, SE = 0.091, t = 2.824, p = .011). Higher T4 G2 externalizing problems significantly predicted higher T5 G2 physical punishment attitudes (B = 0.199, SE = 0.094, t = 2.111, p = .003) and marginally predicted lower T5 G2 warmth attitudes (B = −0.150, SE = 0.091, t = −1.644, p = .100). Finally, higher T2 G1 parental warmth predicted higher T4 G1 warmth (B = 0.257, SE = 0.091, t = 2.824, p = .005). However, the relation between G1 parental warmth and G2 physical punishment attitudes was opposite to the direction predicted; G1 parental warmth predicted higher T5 G2 physical punishment attitudes (B = 0.259, SE = 0.088; t = 2.943, p = .003).
The finding that higher T4 parental warmth was related to higher T5 G2 physical punishment attitudes is surprising from both theoretical and empirical perspectives, given that a large body of literature has demonstrated that parental warmth in adolescence is a potent protective factor (e.g., Pinquart & Gerke, Reference Pinquart and Gerke2019; Rothenberg et al., Reference Rothenberg, Lansford, Bornstein, Chang, Deater-Deckard, Di Giunta, Dodge, Malone, Oburu, Pastorelli, Skinner, Sorbring, Steinberg, Tapanya, Uribe Tirado, Yotanyamaneewong, Alampay, Al-Hassan and Bacchini2020; Shin et al., Reference Shin, Wang, Yoon, Cage, Kobulsky and Montemayor2019). It is important to note that in the model in which T4 parental warmth was included as the sole mediator, it was not significantly related (p = .098) to T5 G2 physical punishment attitudes. Also, the zero-order correlation between T4 parental warmth and T5 physical punishment attitudes was nonsignificant (r = .016, p = .841). Only in the combined mediator model was this relation significant.
Exploratory analyses: G2 age and gender moderation
After correcting for the number of analyses using the false discovery rate, there were no significant differences in the models across G2 age or gender.
Discussion
This is the second experimental study to examine whether a parenting-focused preventive intervention affects parenting attitudes in the next generation. Analyses demonstrated that participation in the Family Bereavement Program (FBP), an intervention for parentally bereaved families, directly reduced G2 physical punishment attitudes 15 years later. The findings also supported a cascade effects model in which intervention-induced improvements in G1 parental warmth at posttest led to fewer G2 externalizing problems and lower anxious romantic attachment in mid-to-late adolescence/emerging adulthood six years later, which in turn led to less favorable G2 attitudes toward physical punishment and more favorable G2 attitudes toward warm parenting, respectively, 15 years after the intervention. The findings did not provide support for a social learning model. Although intervention-induced improvements in G1 parental warmth were maintained six years after the intervention, G1 parental warmth in adolescence/emerging adulthood did not lead to more favorable G2 parenting attitudes in adulthood. We did not find support for the hypothesis that improvements in G2 peer competence or grief would improve parenting attitudes. Below, we discuss the findings in the context of other research on the intergenerational transmission of parenting and their implications for resilience theory and interventions, as well as the study’s limitations and future directions for research.
Direct effects: physical punishment attitudes
Buston et al. (Reference Buston, O’Brien and Maxwell2022) recently called for interventions that can break the cycle of negative parenting. Our findings show that the relatively brief FBP, which focused on improving parent-child relationship quality and effective discipline as well as promoting child competencies, led to less favorable attitudes toward physical punishment. This finding is similar to that of the only other experimental study of whether changes in G1 parenting affect G2 parenting attitudes, which found that a preventive intervention for divorced mothers reduced favorable attitudes toward harsh discipline of emerging adult offspring whose mothers reported using more harsh discipline at program entry (Mahrer et al., Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014). Although parents’ level of physical punishment at program entry might have moderated the relation between the intervention and G2 physical punishment attitudes in the current study, this interaction could not be examined because G1 physical punishment was not assessed at pretest.
Given that more favorable attitudes toward physical punishment are associated with more punitive disciplinary strategies (Azar et al., Reference Azar, Nix and Makin-Byrd2005; Babcock Fenerci et al., Reference Babcock Fenerci, Chu and DePrince2016; Bower-Russa et al., Reference Bower-Russa, Knutson and Winebarger2001; Easterbrooks et al., Reference Easterbrooks, Bureau and Lyons-Ruth2012; Kim & Cicchetti, Reference Kim and Cicchetti2004) and the use of physical punishment is associated with children’s mental and physical health problems (Afifi et al., Reference Afifi, Mota, MacMillan and Sareen2013; Gershoff et al., Reference Gershoff, Goodman, Miller-Perrin, Holden, Jackson and Kazdin2018; Gershoff & Grogan-Kaylor, Reference Gershoff and Grogan-Kaylor2016), this finding suggests that providing a parenting-focused intervention in one generation may positively impact the subsequent generation’s parenting and their children’s mental and physical health problems. The current findings suggest that relatively brief interventions, such as the FBP, may confer even larger public health benefits than previously believed. Should other interventions also show program effects on parenting attitudes, these findings could be used to support funding for such programs, given their long-lasting return on investment.
It is interesting to speculate about how the FBP reduced favorable attitudes toward physical punishment. It is possible that the intervention component focused on discipline, which discouraged use of harsh discipline, such as physical punishment, and taught alternatives to harsh strategies led to decreases in G1’s use of physical punishment. G2s may have endorsed attitudes toward physical punishment that reflected the discipline they experienced after their parents participated in the FBP. It is also possible that self-regulation skills modeled in video skills demonstrations resulted in parents using these skills during interactions with their children, which in turn, led to offspring endorsing attitudes toward parenting that involve using these skills. Further, although not examined in this study, it may be that the child component of the FBP led to improvements in G2 self-regulation which led to less favorable attitudes toward physical punishment. A recent meta-analysis by Robson et al. (Reference Robson, Allen and Howard2020) provided substantial evidence that self-regulation in childhood is significantly related to social competencies, academic performance, and mental health problems in adolescence, all of which have been shown to be related to subsequent parenting in adulthood (e.g., Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009; Raby et al., Reference Raby, Roisman, Fraley and Simpson2015; Shaffer et al., Reference Shaffer, Burt, Obradović, Herbers and Masten2009). Further, self-regulation in adulthood is related to negative parenting, such that parents with poorer self-regulation use more harsh/negative discipline strategies (Bridgett et al., Reference Bridgett, Burt, Edwards and Deater-Deckard2015).
The results highlight the importance of assessing the effects of preventive interventions on both mental health problems and developmental competencies across developmental periods post-intervention. This suggestion echoes a statement made by Coie et al. (Reference Coie, Watt, West, Hawkins, Asarnow, Markman, Ramey, Shure and Long1993) almost three decades ago that understanding the true effects of preventive interventions will require assessments that track participants over development. To date, several randomized controlled trials of relatively brief parenting-focused preventive interventions have demonstrated improved functioning in offspring of participants that last into adulthood, including reductions in mental health problems, physical health problems and substance use problems, less involvement with the criminal justice system, and improvements in competencies, such as academic achievement and work competence (Brody et al., Reference Brody, Yu, Miller, Ehrlich and Chen2019; Herman et al., Reference Herman, Mahrer, Wolchik, Porter, Jones and Sandler2015; Sandler et al., Reference Sandler, Tein, Wolchik and Ayers2016b, Reference Sandler, Gunn, Mazza, Tein, Wolchik, Kim, Ayers and Porter2018; Spoth et al., Reference Spoth, Trudeau, Shin and Redmond2008, Reference Spoth, Redmond, Shin, Trudeau, Greenberg, Feinberg and Welsh2022; Wolchik et al., Reference Wolchik, Sandler, Tein, Mahrer, Millsap, Winslow, Vélez, Porter, Luecken and Reed2013, Reference Wolchik, Tein, Sandler and Kim2016, Reference Wolchik, Tein, Winslow, Minney, Sandler and Masten2021). Along with the findings of Mahrer et al. (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014), the current findings suggest that relatively brief parenting-focused interventions may have direct and cascade effects on additional domains of functioning.
Cascade effects
Physical punishment attitudes
Decreases in externalizing problems emerged as a significant mediator of the cascade effects of the FBP to lower physical punishment attitudes in the combined model. The FBP improved G1 posttest parental warmth, which led to decreases in G2 externalizing problems at the 6-year follow-up, which led to lower G2 physical punishment attitudes. The link between parental warmth and externalizing problems in mid-to-late adolescence/emerging adulthood is consistent with numerous non-experimental studies (Rothenberg et al., Reference Rothenberg, Lansford, Bornstein, Chang, Deater-Deckard, Di Giunta, Dodge, Malone, Oburu, Pastorelli, Skinner, Sorbring, Steinberg, Tapanya, Uribe Tirado, Yotanyamaneewong, Alampay, Al-Hassan and Bacchini2020). The link between higher externalizing problems in adolescence and more favorable attitudes toward physical punishment in adulthood is consistent with the findings of studies on the intergenerational transmission of parenting (Capaldi et al., Reference Capaldi, Pears, Patterson and Owen2003; Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009; Simons et al., Reference Simons, Whitbeck, Conger and Wu1991). The relation between externalizing problems in adolescence and later attitudes toward physical punishment is likely due, in part, to the continuity of externalizing problems and the aggression-supporting cognitive styles associated with them (duBow et al., Reference duBow, Huesmann and Boxer2003; Petersen et al., Reference Petersen, Bates, Dodge, Lansford and Pettit2015; van der Ende et al., Reference van der Ende, Verhulst and Tiemeier2020).
Warmth attitudes
Reductions in anxious romantic attachment in adolescence/emerging adulthood mediated the cascade effects of the FBP on warm parenting attitudes. Specifically, the FBP led to improvements in G1 posttest parental warmth, which led to decreases in G2 anxious romantic attachment at the 6- follow-up, which led to more favorable G2 attitudes toward parental warmth. To our knowledge, this is one of the few prospective studies that include multiple developmental periods to show that romantic attachment affects parenting attitudes or behaviors (see Labella et al., Reference Labella, Raby, Martin and Roisman2019 and Shlafer et al., Reference Shlafer, Raby, Lawler, Hesemeyer and Roisman2015 for exceptions). This finding is consistent with a review of more than 60 studies on the links between self-reported adult attachment styles and parenting which found that romantic insecurity is related to less sensitive, supportive, and responsive parenting behaviors (Jones et al., Reference Jones, Cassidy and Shaver2015).
The association between less romantic anxiety and higher warmth attitudes may be due to differences between securely and insecurely attached individuals’ focus on their own distress and attachment needs, which affects the mental resources needed to respond accurately to another’s needs or the strong desire for closeness, support, and love associated with attachment anxiety. The focus on one’s own attachment needs may taint caregiving motives with desires for acceptance and gratitude, which impair responsiveness (Mikulincer & Shaver, Reference Mikulincer and Shaver2012). Alternatively, Collins et al. (Reference Collins, Ford and Feeney2010) suggest that the discomfort with emotional expression or difficulty regulating one’s own emotions associated with attachment insecurity may explain why responsive caregiving for offspring in distress might be particularly difficult for those who have high levels of romantic attachment anxiety.
The FBP did not have direct or indirect effects on academic competence. Research that shows that parenting interventions led to improvements in academic competence as assessed by objective measures such as GPA and standardized achievement tests (e.g., Brennan et al., Reference Brennan, Shelleby, Shaw, Gardner, Dishion and Wilson2013; Wolchik et al., Reference Wolchik, Sandler, Weiss, Winslow, Briesmeister and Schaefer2007) suggests that the lack of a significant effect in this study could be due to our use of a self-report measure. Academic competence at the 6-year follow-up predicted higher parental warmth attitudes. This finding is similar to that of Mahrer et al.s’ (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) in their study of an intervention for divorced families. Kerr et al. (Reference Kerr, Capaldi, Pears and Owen2009) found that a composite variable of positive adjustment in adolescence that included high school grades predicted constructive parenting in emerging adulthood. The relation between higher academic competence and greater warmth attitudes may reflect the association between higher academic performance and greater educational attainment (Acacio-Claro et al., Reference Acacio-Claro, Doku, Koivusilta and Rimpelä2018). Higher educational attainment is associated with greater commitment to the parenting role (Neppl et al., Reference Neppl, Conger, Scaramella and Ontai2009), greater likelihood of participation in child-rearing educational programs (Haggerty et al., Reference Haggerty, Fleming, Lonczak, Oxford, Harachi and Catalano2002; Harman & Brim, Reference Harman and Brim1980; Johnson et al., Reference Johnson, Harrison, Burnett and Emerson2003; Spoth & Redmond, Reference Spoth and Redmond2000), and authoritative parenting (Dornbusch et al., Reference Dornbusch, Ritter, Leiderman, Roberts and Fraleigh1987), all of which are likely to be related to increased warmth attitudes.
Contrary to our hypothesis, peer competence did not significantly predict later parenting attitudes. The null finding is similar to that of Mahrer et al.s’ (Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014) study of a preventive intervention for divorced families. However, passive longitudinal research has found that social competence is associated with the intergenerational transmission of parenting (Shaffer et al., Reference Shaffer, Burt, Obradović, Herbers and Masten2009). Our results may differ, in part, because of differences in measurement. Shaffer and colleagues assessed parenting quality rather than attitudes toward parenting. Also, Shaffer et al. (Reference Shaffer, Burt, Obradović, Herbers and Masten2009) examined only social competence whereas the current study included other possible predictors.
The findings for G1 parental warmth and attitudes toward parenting were complex. Although intervention-induced increases in warmth showed cascading effects through anxious romantic attachment and externalizing problems to G2 attitudes toward parenting, G1 parental warmth in adolescence/emerging adulthood did not predict G2 attitudes toward warm parenting. This finding is inconsistent with those of the very limited prior research on predictors of parenting attitudes (Mahrer et al., Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014; Thompson et al., Reference Thompson, Hollis and Richards2003). These differences could be due to the use of different measures of positive parenting (Thompson et al., Reference Thompson, Hollis and Richards2003) or the inclusion of a measure of discipline in the model (Mahrer et al., Reference Mahrer, Winslow, Wolchik, Tein and Sandler2014). It was not expected that G1 parental warmth would be significantly related to more favorable G2 physical punishment attitudes. However, given the non-significant correlation between G1 parental warmth and G2 physical punishment attitudes and non-significant direct effect in the single mediator model, we view the significant relation in the combined model as an unstable effect that is most likely a statistical artifact and should not be interpreted as meaningful without replication.
Neither of the two aspects of grief assessed, posttraumatic growth through grief and grief-related social detachment, predicted parenting attitudes. It is possible that other aspects of grief than those tested in the current models may predict parenting attitudes or that the effects of grief do not spread to the domain of parenting attitudes.
Implications
The current study demonstrated both direct and cascade effects of a brief preventive intervention for bereaved families on attitudes toward parenting in the next generation. The experimental design strengthens the inferences that can be made about the intergenerational transmission of aspects of parenting compared to those previously based on retrospective and passive longitudinal designs. Our findings showed that the program affected attitudes toward parenting in the next generation and identified unique pathways that accounted for the program’s effect on attitudes toward physical punishment and warm parenting attitudes.
The current findings add to a growing body of findings from experimental trials that have documented short- and long-term effects of preventive parenting-focused programs on a broad range of adaptation outcomes as well as the results of mediational analyses that have demonstrated that program-induced improvements in parenting-mediated program effects on long-term outcomes through multi-linkage cascade effects (see Sandler et al., Reference Sandler, Ingram, Wolchik, Tein and Winslow2015 for a review). Further, the results of the current study also highlight the power of parenting programs by showing the FBP’s ability to improve an aspect of parenting, attitudes toward parenting, that is likely genetically influenced. Moreover, previous studies have demonstrated that engagement with a preventive intervention, the Family Check-Up, interacted with genetic factors to predict offspring outcomes such that adolescents with genetic risk who were in the program condition were less likely to develop negative outcomes (Elam et al., Reference Elam, Mun, Kutzner and Ha2021; Lemery-Chalfant et al., Reference Lemery-Chalfant, Clifford, Dishion, Shaw and Wilson2018; Shaw et al., Reference Shaw, Galán, Lemery-Chalfant, Dishion, Elam, Wilson and Gardner2019).
This study has important implications for understanding the public health impact of parenting interventions such as the FBP. Given the well-documented relations between parenting attitudes and later parenting behaviors (e.g., Kiang et al., Reference Kiang, Moreno and Robinson2004; Sommer et al., Reference Sommer, Whitman, Borkowski, Schellenbach, Maxwell and Keogh1993) and between positive parenting and children’s mental and physical health problems and competencies (e.g., Hoeve et al., Reference Hoeve, Dubas, Eichelsheim, van der Laan, Smeenk and Gerris2009; Roche et al., Reference Roche, Ahmed and Blum2008), it is likely that the intervention-induced improvements in attitudes toward physical punishment and parental warmth will result in the third generation (i.e., grandchildren) of parents who participated in the FBP experiencing more positive parenting than those of parents who were in the literature control.
These findings extend the breadth of long-term program effects of the FBP. Prior studies have found that the program has long-term effects in emerging/young adulthood to reduce G2 major depression, generalized anxiety disorder, internalizing problems, externalizing problems, suicide thoughts/attempts, and use of mental health services and psychiatric medication Sandler et al., Reference Sandler, Tein, Cham, Wolchik and Ayers2016b, Reference Sandler, Gunn, Mazza, Tein, Wolchik, Kim, Ayers and Porter2018, Reference Sandler, Tein, Zhang and Wolchik2023). The current study identified pathways through which the FBP may have effects that are transmitted to subsequent generations.
This study also adds to the broader literature on resilience theory and preventive interventions as mechanisms for mitigating risk and promoting resilience. The finding that improvements in parenting during childhood/late adolescence had enduring effects on developmental tasks in emerging adulthood is consistent with a large body of research that has identified positive parenting as a key resilience resource (e.g., Baumrind, Reference Baumrind1971; Masten et al., Reference Masten, Burt, Roisman, Obradović, Long and Tellegen2004; Werner, Reference Werner, Goldstein and Brooks2013). The findings also align with resilience theory in emphasizing the dynamic interactions of protective and risk factors across systems and time (Masten, Reference Masten2018), such that promoting positive parenting practices in one generation can have cascade effects across areas of development for offspring and potentially mitigate risk and promote protective resources in subsequent generations. This study also underscores the importance of considering long-term and cross-generational effects in evaluating intervention programs. By examining the complex pathways through which these cascading effects occur, this study adds depth to our understanding of how interventions can leverage resilience processes to foster positive outcomes across generations.
Limitations and future directions
This study should be considered in light of several limitations. First, the dataset did not include a measure of G1 physical punishment. In future research, it would be important to examine the impact of intervention-induced changes in G1 physical punishment as well as parental warmth and the pathways through which intervention-induced changes in physical punishment contribute to G2 parenting attitudes. Second, the dataset did not include measures of G2’s parenting behaviors or their children’s outcomes. Although parenting attitudes are linked to parenting behaviors and parenting behaviors are linked to offspring outcomes, future research should examine intervention effects on G2’s parenting behavior, ideally using gold-standard methods such as parent-child interaction tasks. In addition, testing theoretical models of change wherein intervention effects on attitudes toward parenting lead to improvements in parenting and more positive outcomes in the third generation is a key direction for characterizing mechanisms driving the intergenerational transmission of parenting. Third, academic competence was assessed by self-report rather than more objective measures, such as GPA or standardized achievement tests. Future research should evaluate whether objective measures are more sensitive to change or reliable than self-report measures in this context. Fourth, given the sample size and complexity of the models that were tested, we examined only models that included intervention-induced improvements in parenting. However, testing the role of child competencies, such as coping, as mediators is an important direction for future research. Fifth, although we had a somewhat ethnically diverse sample, we were underpowered to examine ethnicity as a moderating factor. Similarly, we were unable to examine theoretically interesting moderators (e.g., parent gender and cause of parent death) because of small subsamples. Including parent gender and cause of death as moderators in studies with larger samples is an important direction for future research.
Acknowledgments
This research was supported by a grant from the National Institute of Mental Health, R01 MH049155-11A1, Sharlene A. Wolchik’s, Irwin N. Sandler’s, and Michele M. Porter’s work on this paper was supported by a grant from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01HD094334). Jenn-Yun Tein’s work was supported by two grants from the National Institute on Drug Abuse (2R01DA09757; DA09757) and by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01HD094334). Karey L. O’Hara’s work on this paper was supported by a K01 Career Development Award through the National Institute of Mental Health (K01MH120321-01). C. Aubrey Rhodes’ work was supported by a predoctoral fellowship provided by the National Institute on Drug Abuse (T32DA039772).
Competing interests
The authors declare none.
Open access
