Introduction to the Special Section on the Intergenerational Transmission of Risk for Substance Use

Abstract

Six original research papers were submitted to this special section to address questions regarding the intergenerational transmission of risk for cannabis and other substance use. Study teams recruited youth in Iowa, Washington, Oregon, New York, and Arizona in the 1980s–1990s, assessed them into adulthood, and recruited their partners and offspring for another study. All of the studies assessed substance use in two or more generations. Other strengths in this section include the strong representation of fathers, the demographic diversity of the samples as a whole, and the demonstrations of varied statistical and replication approaches. The findings highlight features of parental histories of cannabis use during adolescence that are associated with their children’s risk for cannabis use and factors that explain or weaken intergenerational similarities. Two groups of prevention scholars also offered commentaries on the implications of these studies for prevention and training, and collaboration. It is hoped that the special section will stimulate new hypotheses, replications, and communication among etiological and prevention researchers. Furthermore, the papers highlight that the familial transmission of substance use risk should be taken into account more fully in the design of prevention programs in order to maximize impacts for youth as well as their future offspring.

In 2018, we initiated a call for papers for a special section of Psychology of Addictive Behaviors to encourage the study of the intergenerational transmission of risk for substance use and increase the prevention impact of study findings. In response, intergenerational researchers and prevention experts submitted six empirical papers (see Table 1) and two commentaries and, together, helped to accomplish the aims of the special section. In this Introduction, we summarize the focus and rationale for these aims and highlight some of the emergent properties of the collection of papers that may not be evident in any single paper. We also discuss directions for future study as well as overarching prevention implications.

Table 1.

Characteristics of the Samples and Primary Substance Use Predictors and Outcomes for the Six Studies in the Special Section

State of G2 recruitment	Basis of G2 problem behavior risk	Start of G2 study	Analytic sample size	G3 race/ethnicity	Predictor(s)	Outcome(s)
Study 1: Augustyn, Loughran, Larroulet, Fulco, and Henry, Rochester Youth Development Study-Rochester Intergenerational Study (RYDS-RIGS)
New York	Oversampled males (3:1) and those living in census tracts with high arrest rates for adults	1988, G2 Grade 7 or 8	G2 n = 462 (64% male) G3 n = 462 (50% male)	65% African American 11% Hispanic 14% Mixed or other 10% White	G2 marijuana use trajectories, ages 14–31 years	G3 marijuana use onset to age 25 years
Study 2: Epstein, Bailey, Furlong, Steeger, and Hill, Seattle Social Development Project-The Intergenerational Project (SSDP-TIP)
Washington	Schools from higher crime areas of city	1985, G2 Grade 5	G2 n = 380 (40% male) G3 n = 380 (52% male)	14% African American 11% Hispanic 34% Multiracial 36% White 12% Asian/Pacific Islander 4% Native American	G2 marijuana use trajectories, ages 14–30 years	G3 marijuana, alcohol, and tobacco use; pro-marijuana norms; internalizing/ externalizing symptoms; and grades, ages 6–20 years
Study 3: Kerr, Tiberio, Capaldi, and Owen, Oregon Youth Study-Three Generational Study (OYS-3GS)
Oregon	Schools from neighborhoods with higher juvenile delinquency; boys only	1984, G2 Grade 4	G2 n = 113 (100% male) G3 n = 223 (44% male)	4.5% African American 4.5% Hispanic or Latino 5.8% Multiracial/ethnic 76.2% European American 9.0% Other identities	G2 adolescent polysubstance (ATM) use onset to age 18 years	G3 adolescent polysubstance (ATM) use onset to age 18 years
Study 4: Neppl, Diggs, and Cleveland, Family Transitions Project (FTP)
Iowa	Family residing in rural Midwest following 1980s economic farm crisis; families headedby single mothers oversampled*	1989, G2 Grade 7*	G1 n = 218 (100% female) G2 n = 218 (40% male) G3 n = 218 (54% male)	<1% American Indian <1% Alaskan Native <1% Asian/Pacific Islander <2% Hispanic 98% White non-Hispanic	G1 alcohol problems in adulthood (mean age 40–42 years G2 alcohol and marijuana problems, age 21 years	G2 alcohol and marijuana problems, age 19 years G3 aggression, ages 6-10 years
Study 5: Rothenberg, Sternberg, Blake, Waddell, Chassin, and Hussong, The Adult and Family Development Project (AFDP)
Arizona	50% had ≥ 1 parent with AUD; 50% non-AUD matched control	1988, G2 ages 10.5-15.5 years	G1 n = 454 (45% male) G2 n = 454 (53% male) G2 = 298 (53% male) G3 = 399 (51% male)	58.2% White/Not Hispanic 30.3% Hispanic 1.3% Native American 1.8% African American 8.5% Other	G1 cannabis use history and AUD in adulthood (mean age ~40 years) G2 cannabis use history and AUD in adulthood (mean age ~40 years)	G2 late adolescent to early adulthood cannabis use G3 late adolescent to early adulthood cannabis use
Study 6: Tiberio, Kerr, Bailey, Henry, and Capaldi, 3 Generational Research Consortium (3GRC), including RYDS-RIGS, SSDP-TIP, OYS-3GS
New York, Oregon, and Washington	See entries above for RYDS–RIGS, SSDP–TIP, OYS—3GS	See entries above for RYDS—RIGS, SSDP—TIP, OYS—3GS	G2 n = 1081 (64% male) G3 n = 1017 (49% male)	38.3% African American 10.9% Hispanic or Latino 15.3% Multiracial/ethnic 32.8% European American 5.8% Other identities	G2 adolescent cannabis use onset to age 18 years	G3 adolescent cannabis use onset to age 18 years

^*108 families headed by a single mother added to the sample in 1991 when G2 were in Grade 9.

Note. G1, G2, G3 = Generation 1, 2, and 3. AUD = Alcohol Use Disorder. ATM = alcohol, tobacco, and marijuana. The terms cannabis and marijuana are used here according to the language used in each report. The first author compiled this information from the above papers and from prior papers on these samples; any errors are his.

Focus of the Special Section

Intergenerational designs can be used to illuminate the extent to which parents’ and their children’s substance use patterns are related, the mechanisms that explain any associations, and the circumstances under which these associations are stronger or weaker. Researchers infer that similarities between parent and child behavior reflect intergenerational transmission if assessment of parents’ behavior significantly predates that of children. Statistical modeling also is used to rule out other continuities (such as in socioeconomic factors) as explanations or consider them as transmission mechanisms. Of special interest, presently, are associations that parents’ substance use during their own adolescence—in most cases predating the birth of their children—may have with their children’s substance use. Extending Bornstein, Putnick, and Esposito’s (2017) discussion of terminology in developmental science to the intergenerational context, one could describe these associations in terms of continuity (similar mean levels in the parent and child generations), stability (similar rank ordering of parent and child generations relative to their peers), or other parallels across generations in substance use patterns or changes over time (e.g., congruence; Kerr, Tiberio, Capaldi, & Owen, in press).

The intergenerational transmission of substance use is relevant to prevention because it is assumed to occur at least partially through modifiable biopsychosocial mechanisms (vs. non-modifiable genetic risks only, for example). Thus, intergenerational studies may help prevention scientists identify and target higher risk individuals, sensitive periods of life (e.g., early adolescence), and particularly consequential patterns or types of substance use. Findings on intergenerational discontinuity and instability (e.g., why some adolescents use at lower rates than their parents did) also may yield new insights about risk and protective factors. Additionally, whereas traditional longitudinal studies show the consequences of adolescents’ substance use in their own lives, intergenerational studies illuminate how such consequences may affect their families of procreation—thus demonstrating the potential scope of prevention effects.

We attempted to maximize the contributions the series of papers would make to this field by shaping the focus of the special section in the following ways. First, submissions of papers were solicited based on studies that had assessed at least two generations of individuals prospectively across adolescence. As has been described in the special section papers and elsewhere (Thornberry, 2016), such study designs have a number of qualities that enhance validity and avoid bias when tracing the development and intergenerational transmission of risk for substance use. The strong showing of the multiple prospective intergenerational studies that contributed to the special section reflects many years of commitment and work both by the research teams and funding agencies. These studies are coming to fruition for addressing questions regarding intergenerational transmission of substance use, now that most of the third generation are well into their adolescent years. A second way we attempted to enhance the impact of the special section was to invite studies with a focus on adolescents’ cannabis and polysubstance use—given the increased interest in the causes and consequences of these behaviors and given that the emphasis of previous intergenerational studies of substance use more often has been on alcohol and tobacco use (Handley & Chassin, 2009; Kandel & Wu, 1995; Kerr, Capaldi, Pears, & Owen, 2012). Third, researchers were encouraged to submit papers addressing questions of intergenerational discontinuity or instability as well as continuity or stability, given the potential of such research to identify prevention targets. Several studies in this issue investigated factors that could be associated with disruptions in parent–child transmission of risk for substance use, including psychological processes (e.g., active coping during adolescence; Rothenberg et al., in press) and demographic characteristics (e.g., same gender parent–child dyads; Tiberio, Kerr, Bailey, Henry, & Capaldi, in press), whereas one study considered parent and child behaviors that help explain intergenerational stability in substance use (Neppl, Diggs, & Cleveland, in press). Still other studies in this series considered which developmental patterns of substance use were more or less durable across generations (e.g., polysubstance use, adolescent-onset cannabis use; Epstein, Bailey, Furlong, Steeger, & Hill, in press; Kerr et al., in press; Tiberio et al., in press). Finally, an important element of the call for papers was a request for greater attention to the prevention implications of the findings of intergenerational studies than is typical in etiological research reports. We encouraged this prevention emphasis by explicitly asking the investigative teams to speculate further than they normally would regarding prevention issues and by inviting two groups of substance use prevention scholars (Etz, Goldstein, Lopez, & Blanco, in press; Haggerty & Carlini, in press) to offer prevention-oriented commentaries.

Emergent Properties of the Collective Papers

Each paper in the special section makes an independent contribution to the field. Yet, we also wish to highlight features of the special section that may not be apparent when considering any individual study but that emerge when reflecting on the project as a whole.

Advanced methodological and statistical approaches.

A notable feature of the papers in this special section is the methodological sophistication required to answer questions about substance use using intergenerational designs. Challenges routinely faced in longitudinal research on child adjustment are magnified in intergenerational studies. For example, although the parent-generation participants in these studies were from a similar cohort, the offspring generation were born across a prolonged period. Even if one considers only the peak child-bearing years for the parent generation, that period spans about 20 years. Thus, the lengthy follow-up period needed to assess their offspring during adolescence is a great obstacle. Additionally, the contributing authors often had to grapple with cohort effects, low prevalence outcomes, selection biases, nesting of individuals in families, and the challenge of parsimoniously describing complex patterns across multiple observation years. Four of the papers used latent modeling approaches to identify groups of individuals who were similar in terms of onset or patterns of substance use across time; these approaches included group-based trajectory modeling (Augustyn, Loughran, Larroulet, Fulco, & Henry, in press), growth mixture modeling (Epstein et al., in press), and two forms of discrete-time survival mixture analyses (Kerr et al., in press; Tiberio et al., in press). The other two studies used ordered logistic regression (Rothenberg et al., in press) and structural equation models (Neppl et al., in press) to capitalize on the long-term prospective assessment designs of their studies that permit testing of associations among temporally sequenced variables. For example, Neppl and colleagues considered child outcomes associated with their parents’ and grandparents’ substance use—carefully specifying pathways through parents’ middle and late adolescence and early adulthood and continuing through their children’s early and middle childhood. As a group, the six papers serve as useful illustrations of how to apply these statistical approaches to complex datasets, and they will inspire the use of these methods in future studies of intergenerational transmission of substance use and other phenomena (e.g., parenting, depression, violence).

The value of replication.

There are growing concerns in the behavioral and social sciences regarding the reliability and validity of findings (Camerer et al., 2018). The relatively small sample sizes of many community studies, including prospective intergenerational studies (which are intensive and expensive) raise concerns about the replicability of findings. Furthermore, the methods used to analyze complex longitudinal data are often sophisticated but statistically underpowered (e.g., identifying small latent classes in a moderately sized sample; predicting low-frequency outcomes; testing moderation of modest effects). Thus, this special section offers a valuable forum for replicating important questions regarding the intergenerational transmission of substance use risks. Communication over time among many of the teams that contributed papers to this special section has led to some similarities in measures used in the studies, particularly for the intergenerational aspects, with a view to facilitating later collaborative analyses. In addition, when the research groups were working on relatively similar questions, they often cross cited and built on each other’s work. For example, findings from the Seattle Social Development Project–The Intergenerational Project (SSDP–TIP) have been used to justify hypotheses in the Rochester Youth Development Study–Rochester Intergenerational Study (RYDS–RIGS). Also, the author teams have speculated on whether, for example, discrepant findings from RYDS–RIGS and the Oregon Youth Study–Three Generational Study (OYS–3GS) were explained by differences in sampling, operational definitions, developmental period of focus, or analytic approach. Many of the author groups also have contributed to special journal issues on intergenerational continuities in other constructs of shared interest, such as antisocial behavior and parenting (e.g., Conger, Belsky, & Capaldi, 2009). Thus, a number of factors facilitated replications in the studies represented here.

Multiple forms of replication.

An unexpected property of the collective papers is the demonstration of the many forms replication can take in intergenerational studies. First, Rothenberg and colleagues (in press) completed a within-study, cross-generational replication by testing models of risk transmission and moderation for parents and adolescents (G1–G2) and then examining the same models for those adolescents who became parents of adolescents (G2–G3). Second, in their data synthesis, Tiberio and colleagues (in press) combined three intergenerational datasets to examine similarities in cannabis onset timing during adolescence. In doing so, they tested the similarities and differences among the studies in terms of the cannabis onset patterns for the parent and offspring generations, as well as the cross-study similarity in predicting offspring onset from parents’ adolescent onset. Data harmonization approaches, involving post hoc transformations in variable scoring to provide comparable measurements across studies, allow for analyses with larger sample sizes and have been utilized with large-scale epidemiological studies (e.g., Healthy Obese Project; Doiron et al., 2013). As discussed by Hussong, Curran, and Bauer (2013), the advantages of simultaneous analysis of raw data pooled from multiple studies include increased power due to larger sample size and the chance to build a more cumulative science. Third, the Rothenberg et al. (in press), Kerr et al. (in press), and Tiberio et al. (in press) papers evaluated replication across historical cohort; for example, Kerr and colleagues (in press) found that, consistent with national studies such as Monitoring The Future (Johnston et al., 2020), the offspring generation showed later onset on alcohol and tobacco use, but not cannabis use, than the parent generation had. Finally, the Epstein et al. (in press) and Augustyn et al. (in press) studies together provide a clear example of a constructive replication (Lykken, 1968) in that they used different statistical methodologies with different samples to answer similar research questions; specifically, they estimated classes based on parents’ cannabis use histories and then predicted offspring outcomes, including cannabis use, from these classes. These studies add further nuance to prior intergenerational work on the implications of parents’ cannabis use in adolescence versus emerging adulthood for their children’s substance use risk (e.g., Knight, Menard, & Simmons, 2014). Thus, the studies in the special section provide models of varied replication approaches that can be used to strengthen the knowledge base on intergenerational transmission and to generate new hypotheses.

Generalizability of findings.

Although replication of scientific findings is valuable, replication may not mean that findings generalize across important populations. On its own, a single intergenerational study may be unable to rule out potential threats to the external validity of their findings, given key sampling issues related to geographic region, urbanicity, parent gender, racial/ethnic homogeneity, and the over- or under-representation of higher risk individuals (Capaldi, Kerr, & Tiberio, 2018). This point is illustrated in Table 1, which lists sample characteristics for each study in the present special section. It is apparent in this table that each sample has at least some limitations in terms of potential generalizability. However, taken as a whole, the collective studies represent a broader and more diverse population than does any individual intergenerational study. In addition, given that several patterns were consistent across the studies (e.g., Augustyn et al., in press; Epstein et al., in press) or were tested across studies after accounting for sample differences (Tiberio et al., in press), there is heightened confidence that the models of risk transmission apply to a diverse range of community and family contexts.

The challenge and value of considering fathers.

Another emergent property of the special section was that fathers were well represented in the contributing studies, comprising more than one half of the parent generation in four of the six papers (Augustyn et al., in press; Kerr et al., in press; Rothenberg et al., in press; Tiberio et al., in press) and participating at high rates in the others. Although the authors often do not draw attention to this point, we wish to highlight why the inclusion of fathers in sufficient numbers to represent them or to make comparisons with mothers is unusual and therefore a major contribution.

In research on child development, tests of fathers’ influences often are either obstructed by, or complicated by, a host of cultural, pragmatic, and scientific issues. A first set of obstacles relates to fathers’ reticence to participate and researchers’ limited motivation to insist on it, given assumptions about fathers’ incremental value (e.g., as a source of influence; as an informant) and limited study resources to recruit and retain them when the child’s mother is participating. The studies for this special section successfully navigated these issues, in part because they originally included boys as focal participants. The relationships the study teams had with participating boys that began long before they became fathers facilitated their recruitment for the follow-up intergenerational studies. Second, the theoretical models underpinning the studies presented here require an accounting of the parents’ developmental histories and behaviors during their children’s lives (as well as data on the second parent, preferably), and these rich prospectively measured histories are available on the fathers in these studies. Finally, many of the research teams (e.g., Neppl et al., in press) have a tradition of utilizing multimethod, multi-informant measurement models wherein fathers’ reports contribute to the robustness of latent constructs. Perhaps this helps explain why the teams that contributed to this special section were motivated to prioritize the retention of both fathers and mothers and were so successful at doing so.

A second set of challenges that tends to limit consideration of fathers in etiological studies on children concerns fathers’ physical absence from their children’s lives. Father–child contact decreases across children’s development as parental relationships erode, and father involvement can be partial, intermittent, or absent and can occur in the contexts of shared custody or nonresidential arrangements. Father contact also is predicted by fathers’ early transition to parenthood and is more sensitive to their other characteristics and life circumstances than is the case for mothers (i.e., father involvement patterns are nonrandom; see Thornberry, 2005). Thus, intergenerational research teams must decide how to conceptualize less than fully present fathers as informants on child behavior and as direct and indirect influences on their children. For example, some transmission pathways (e.g., modeling; monitoring) may depend on fathers’ physical presence, whereas others (e.g., genetic; stability of family contextual disadvantage) may not.

The contributing authors handle these challenges in various ways, such as by designating some degree of father contact as an eligibility criterion for study participation or for consideration as an informant. Several studies control for parent gender, which may help correct bias associated with fathers’ versus mothers’ nonparticipation in the intergenerational follow-up studies. Still, others control for father contact in study models and for family characteristics associated with father absence. Although not pursued in the studies here, in prior work, several of the teams have considered father contact as a moderator of associations between fathers’ and children’s behaviors (Kerr, Tiberio, Capaldi, & Owen, 2020; Thornberry, Freeman-Gallant, & Lovegrove, 2009; Thornberry, Krohn, & Freeman-Gallant, 2006). These decisions have a number of tradeoffs. For example, families with completely absent fathers are less represented in some of the studies. Another example of a tradeoff is that if boys’ substance use predicts their future low involvement as a father, then controlling for this involvement diminishes the estimated magnitude of father–child risk transmission (Schisterman, Cole, & Platt, 2009). Still, these decisions reflect the careful attention of these researchers to these thorny questions. The possible effects of such design decisions can be probed in robustness analyses and replications.

Future Directions and Open Questions

A number of priorities for future research are raised by the authors of the papers in this special section. These include, first, that risk for offspring substance use from their parents’ adolescent and adult use of cannabis and other substances needs further examination, including associated mechanisms. For example, parents’ cannabis use during adolescence may have an influence on offspring via detriments in social capital (Curran, 2007) for the family of procreation or be partially or completely mediated by parents’ adult cannabis use. Relatedly, given associations between parents’ histories (e.g., Kerr et al., in press), tests of how the risk from parents’ substance use during adolescence is amplified or offset by their partners’ use in adulthood would be worthwhile. Second, additional consideration of intergenerational congruence in timing, patterns, and severity of substance use is warranted in order to understand factors underlying such congruence and to identify how to shape questions regarding intergenerational mechanisms of continuity and discontinuity. Third, attention toward gender differences in mechanisms of familial transmission may illuminate models of etiology and be relevant to prevention development.

Fourth, although most of the studies here focused on cannabis use by parents and children, further investigation is needed of the extent to which these pathways reflect transmission of risk for problem behavior more broadly (e.g., including antisocial behavior), or substance use more generally (e.g., including alcohol use), or are specific to cannabis use. Nadel and Thornberry (2017) considered this idea in their study of homotypic (e.g., parent substance use to child substance use) versus heterotypic (e.g., parent substance use to child problem behavior) continuity in the RYDS–RIGS sample; notably, they found the patterns depended on parent gender. Clarification on these critical issues will indicate the extent to which prevention aimed at delaying cannabis onset by addressing an underlying problem behavior syndrome would be improved if cannabis-specific programming were added. A fifth priority for future intergenerational studies is the consideration of more serious outcomes—such as substance use disorders that, in part because of the relative immaturity of the third-generation participants, have not yet been considered. Increased knowledge in all these areas will inform and sharpen the focus of preventive interventions.

Another issue for further consideration relates to the historic periods the contributing studies represent. As shown in Table 1, the original focal participants (Generation 2 parents) in the six studies were born in the mid-1970s and expected to graduate from high school in the early to middle 1990s. This similarity is no coincidence, as the original projects were funded by the National Institutes of Health during a period of special interest and support for longitudinal studies in this area. Furthermore, in many cases the originating principal investigators (e.g., Drs. Gerald Patterson, Rand Conger, Terence Thornberry, Laurie Chassin, and David Hawkins) communicated regarding study designs and were influenced by each other’s work. Although the relative narrowness in historic timing of the Generation 2 parent cohort is advantageous from the standpoint of replication, it also imposes some limitations. For example, the landscape of substance use has changed significantly in the years since these parents were adolescents. Specifically, the prevalence of alcohol use is dramatically lower among present-day adolescents, whereas the prevalence of cannabis use (which was at a relative low point in the early 1990s) is higher (Johnston et al., 2020). If cannabis use was more strongly associated with other deviant behavior in the 1990s than it is today, then it is possible that the intergenerational continuities observed here were driven by the higher risk adolescents in the parent generation. Additionally, national and regional liberalization trends complicate the picture, as cannabis use is now legal for medical and nonmedical (recreational) purposes in several of the states where the study participants reside. Furthermore, the potency of cannabis approximately tripled from 1995 to 2014 (ElSohly et al., 2016). Thus, new studies of modern cohorts and comparisons among cohorts within Generation 3 samples will be needed to build on the knowledge emerging from this special section. Indeed, intergenerational studies, including by teams who contributed to the present special section (e.g., Kosterman et al., 2016), already are helping the field grapple with how shifts in the causes, contexts, and consequences of substance use affect familial risk transmission.

Conclusions

This special section on the intergenerational transmission of substance use pools the expertise of numerous theorists, methodologists, and prevention experts and makes use of remarkably rich datasets in which investigators, staff, funding agencies, and participating families have invested for decades. Strengths of the studies include designs that covered two or more generations of prospectively assessed individuals, an unusually high level of participation by fathers and mothers, the demographic diversity of the samples as a whole, the showcasing of innovative statistical and replication approaches, and the emphasis on prevention implications.

Findings from the studies in this special section broaden understanding of parental histories of substance use during adolescence as risk factors for their offspring’s substance use. Thus, risk behaviors in adolescence may cast very long shadows for the next generation. In terms of prevention implications of the notion that substance use during adolescence shows some intergenerational stability, one can consider two different vantage points. One perspective on the intergenerational association concerns its implications for the child and how the parent and the child’s environment are operating in the present moment to affect risk and protection. From this standpoint, parents’ histories are not modifiable but could be used to identify high-risk populations, specify and intervene on mechanisms by which parents’ histories would otherwise increase risk during the child’s lifetime, and ultimately—as Haggerty and Carlini (in press) note—avoid the trap of self-fulfilling prophecy. However, another vantage point on intergenerational continuities centers on parents when they were still adolescents and the long-term negative impacts their early involvement in cannabis use could have on their later adjustment and family formation. From this standpoint, the assembled studies have a different set of prevention implications that are proactive rather than reactive; namely, underscoring the importance of delaying cannabis use or mitigating its effects on late adolescent and early adulthood adjustment—ideally, long before the transition to parenthood.

In closing, the empirical papers included in this special section offer theoretical and methodological advances in the study of familial transmission of risk for substance use. The commentaries also highlight new connections, training priorities, translation opportunities, and make valuable recommendations for prevention. We are hopeful these papers will stimulate further inquiry, replication, and collaboration in etiological and prevention research.