Table of Contents | Previous | Next |
Chapter Four: Review of Research on the Causes, Correlates, and Consequences of Romantic Relationships in Adolescence
Despite theoretical interest in the role of adolescent romantic experiences in adult development, up to now, research on adolescent experiences of intimacy has focused primarily on sexual behavior, often to the exclusion of the relationships in which those behaviors take place (Brown, Feiring, and Furman, 1999; Furman and Shaffer, 2003; Giordano, 2003). Over the past few years, several important edited volumes (Crouter and Booth, 2006; Florsheim, 2003b; Furman, Brown, and Feiring, 1999) have drawn attention to the need for research on adolescent relationships per se, and their role in adult development in particular. Even the most recent of these volumes, however, ends with the observation that claims about the continuity of romantic experiences between adolescence and adulthood continue to rest more on speculation than on empirical evidence (Kan and Cares, 2006).
Yet research on the causes, correlates, and consequences of adolescent romantic relationships does exist, albeit scattered across a wide range of academic disciplines and publishing outlets. The goal of this chapter is to assemble this literature in a single place and to take stock of the accomplishments of work in this area to date. In so doing, the chapter offers three potential benefits for efforts to understand the role of adolescent romantic experiences in adult development. First, establishing a common base of knowledge on adolescent romantic relationships will assist researchers in designing new studies that build upon and refine what has already been established. Second, by organizing the existing research within a single framework, this review highlights directions for future research on propositions and associations that have been understudied or overlooked entirely. Third, a review of the field to date allows interventions targeting adolescent romantic relationships to proceed from the firmest possible empirical foundation and promotes the accumulation of future empirical research in ways that maximally inform policies and programs.
The Scope and Procedures of This Review
In general, the review was designed to examine research relevant to understanding the role of adolescent romantic relationships in adult development. This broad aim directed our attention toward studies addressing the integrative framework described in Chapter Three. It should be noted at the outset that no study has brought together all parts of the framework in a single investigation. Instead, this review assembles research that has examined each of the three broad groups of variables identified within the framework. First, we identified studies linking relationship behaviors and experiences during adolescence to antecedent conditions, e.g., variables assessed in the family of origin and aspects of the social context. Second, we identified studies examining variables likely to be associated with relationship experiences during adolescence, e.g., attitudes and beliefs, peer group membership, physical development. Third, we identified studies examining the consequences of experiences in adolescent romantic relationships for outcomes assessed during adolescence and later adulthood.
For each of these categories of research, we identified relevant studies in three ways. First, we consulted prior reviews of research on adolescent dating and romantic relationships (e.g., Bouchey and Furman, 2003; Collins and Madsen, 2006; Furman and Shaffer, 2003; Giordano, 2003). Although each of these reviews identified important studies in this area, none attempted to be systematic or comprehensive and none was specifically designed to understand adolescent romantic relationships as precursors of adult marriages. Second, to identify studies not included in prior reviews, we conducted literature searches through Internet-based databases (i.e., Web of Science, PsychInfo, Social Science Abstracts) using a wide range of keywords (e.g., adolescence, adolescent, romance, intimacy, relationships, etc.). Abstracts for new studies identified in this way were scanned for their relevance to the goals of the current project. Third, we consulted with colleagues in this area and solicited recommendations of additional relevant research from them. This combination of procedures offers confidence that the studies assembled here are reasonably comprehensive of the existing work in the area to date.
Once we had identified a potentially relevant study, we based the decision to include it in the review on several considerations. First, all studies included here in some way address romantic relationships explicitly. As has been noted, a vast array of research on adolescence has examined sexual behaviors in this population without addressing the relational context of those behaviors directly. That research has been the topic of several reviews of its own (e.g., Bissell, 2000; Bonell, 2004; Brooks-Gunn and Paikoff, 1997; Card, 1999; DiIorio, Pluhar, and Belcher, 2003; Herold and Marshall, 1996; Santelli et al., 2006) and so was not included here. However, we did include research on sexuality and sexual behaviors that also included analyses of relationships per se.
Second, the studies included in the review had to make some attempt to explain or account for variability in individual outcomes. Studies that simply described adolescent relationships were reviewed in Chapter Two of this report and are not reviewed again here. Case studies, ethnographies, and studies of very small samples were generally excluded by this criterion. However, studies that offer descriptive data and also attempt to account for variability in outcomes are reviewed in both chapters.
Third, this review defined adolescence as the period of life from 10 to 19 years of age, consistent with the definition of adolescence put forth by the World Health Organization. As a result, research on romantic relationships among college students was, for the most part, excluded from the review. The only exceptions to this rule were cases in which the only available study on a topic was one that addressed college students or when the study explicitly stated that all relationships were assessed among respondents under age 19. These exceptions are noted in the text.
Fourth, the studies included in the review come from peer-reviewed journals, edited book chapters, or reports from reputable research organizations. Unpublished doctoral dissertations were not included.
Fifth, our inclusion criteria did not include any restrictions on the age of the study, but the text emphasizes more recent work over older work.
Sixth, our inclusion criteria allowed research on same-sex relationships, but research in this area is so underdeveloped (Diamond, 2003; Savin-Williams, 2001) that the final pool of studies included very few that addressed same-sex relationships.
An important consideration throughout the conduct of this review was whether the studies addressed low-income populations specifically. It bears repeating that the broad goal of this report is to examine research relevant to promoting healthy marriages within low-income populations specifically. Yet, as has also been mentioned previously, there is almost no research that directly addresses the precursors of healthy marriages in low-income populations—and even less research on adolescent relationships in low-income populations. For these reasons, the income level of the sample was not a criterion for inclusion in this review. Instead, the review addresses research from a broad range of samples, from sample of convenience to nationally representative samples, with studies of nationally representative samples receiving more emphasis in the text. Throughout, the analyses of the research speculate on the extent to which the results of existing research are likely to generalize to low-income populations.
Comments on Methodological Issues
Interpretations of the results of research on adolescent romantic relationships must necessarily be qualified by the limitations of the research methods used to obtain those results. Thus, it makes sense to begin the literature review by commenting on methodological issues that pertain to the majority of the studies reviewed here. These methodological issues then provide a context for the analyses of the substantive findings that follow.
Lack of Agreement on Definitions of Key Terms
In Chapter One, we noted that researchers have yet to reach a consensus on what makes an intimate relationship healthy, let alone what makes intimate relationships healthy among adolescents. The challenge of defining relevant terms concretely affects other areas of this field as well (Kan and Cares, 2006). For example, research on adolescent romantic relationships has yet to provide a concrete definition of what a romantic relationship is and what it is not. As noted in Chapter One, researchers tend to leave the task of definition to study participants, who are free to decide for themselves whether and when they are in romantic relationships. Analyses of data from the Add Health study reveal the potential consequences of this practice. In Add Health, respondents were asked to nominate up to three individuals with whom they had experienced a romantic relationship. Because Add Health examines school-based samples, many of the individuals so nominated were also respondents, allowing analyses of the extent to which nominations were reciprocated. In fact, a large percentage of the nominations were not reciprocated, i.e., the two individuals did not agree on whether or not they had ever been in a romantic relationship (Carver and Udry, 1997; Kennedy, 2005). If the definition of a relationship is so vague that two individuals so frequently disagree on whether they had one, the looseness of existing definitions of terms, such as dating, going steady, falling in love, etc., raise serious questions about what research on these phenomena is actually studying.
Lack of Data from Representative Samples
As noted in Chapter Two, at least five studies of nationally representative samples include assessments of variables relevant to understanding adolescent experiences in romantic relationships. Yet, as also pointed out in Chapter Two, to date analyses of these data have generally been descriptive. Theory-testing within these data sets—and even analyses that compare youth across levels of SES, race, and ethnicity—is possible, but so far it has not been conducted. Thus, research that tests hypotheses about the causes and consequences of adolescent romantic experiences has tended to rely on data from smaller samples and samples of convenience. This limitation particularly affects the ability of existing research to comment on low-income youth.
Lack of Longitudinal Data
To best understand how specific relationship experiences during adolescence may account for outcomes during later life, researchers have been unanimous in calling for a longitudinal approach (e.g., Collins and Van Dulmen, 2006; Eccles and Gootman, 2002; Furman and Shaffer, 2003; Giordano, 2003; Kan and Cares, 2006). Yet longitudinal data on adolescents’ romantic relationships have accumulated slowly. Even within theory-based research on small samples, most research on the correlates and consequences of adolescent relationships to date has drawn from cross-sectional or retrospective data collected at a single assessment, with the Add Health study being a notable exception. The weaknesses of cross-sectional designs for drawing inferences about causal influence have been well documented (e.g., Brewin, Andrews, and Gotlib, 1993; Karney and Bradbury, 1995). In addition to the Add Health study, longitudinal research that has assessed adolescents’ relationship experiences directly include the Toledo Adolescent Relationships Study (e.g., Giordano, Longmore, and Manning, 2006; Giordano, Manning, and Longmore, 2005), the Minnesota Longitudinal Study of Parents and Children (e.g., Carlson, Sroufe, and Egeland, 2004; Sroufe et al., 2005), a study of 103 German middle-school children (Seiffge-Krenke, Shulman, and Klessinger, 2001), and the Oregon Youth Study (e.g., Capaldi and Clark, 1998). These studies offer suggestive evidence for the continuity between adolescent and adult romantic experiences, but all three studies have sampled predominately white, middle-class families, leaving the generalizability of their results to other populations (e.g., low-income populations) an open question.
Failure to Acknowledge Age Differences Within Adolescence
Researchers commonly define adolescence as a stage of life spanning from approximately the ages of 10 to 19. This time is of particular interest to those studying romantic relationships because, as many have noted, the capacity for relationships changes drastically during this period (e.g., Brooks-Gunn, Petersen, and Eichorn, 1985; Brown, 1999, 2006; Collins and Madsen, 2006; Erikson, 1968; Laursen and Collins, 1994; Sroufe et al., 2005; Sullivan, 1953). For example, the cognitive skills that permit an individual to feel empathy and understand the perspective of another person do not mature until later in adolescence (e.g., Hill and Palmquist, 1978). Older teens use different strategies to resolve interpersonal conflicts than do younger teens (Feldman and Gowen, 1998). The meanings ascribed to love and intimacy change significantly over this period as well (e.g., Bouchey and Furman, 2003; Connolly et al., 1999; Waldinger et al., 2002).
To the extent that adolescents of different ages are differentially prepared for romantic relationships, the same experiences occurring at different ages may have different associations with later young adult outcomes. Experiences that may have positive consequences for development when they occur later in adolescence (e.g., falling in love, exclusivity, commitment, spending significant amounts of time together) may have negative consequences when they occur too early. Existing research on these issues has rarely acknowledged this possibility. Instead, most longitudinal work comparing different stages of development either groups respondents within a stage regardless of age or treats age as a control variable. As a consequence, the way that age may affect the links between adolescent romantic experiences and adult outcomes remains an open question.
Reliance on Data from Individuals Rather Than Couples
An irony of research on romantic relationships is that the great majority of it gathers data solely from individuals (Furman, 1984; Karney and Bradbury, 1995). A reliance on individual reports similarly characterizes research on romantic relationships among adolescents (Collins and Madsen, 2006; Furman and Hand, 2006). As a result, several important questions about these relationships have been overlooked within this literature. Although it is hard to imagine that adolescent relationships are not affected in significant ways by characteristics of partners, “In the study of adolescent romantic relationships, no published findings have yet addressed questions of the partner’s identity or the impact of the partner’s characteristics on relationship quality” (Collins and Van Dulmen, 2006, p. 80). Without such data, it remains possible that the consequences of relationship experiences during adolescence stem entirely from attributes of the partner (e.g., age difference, substance use, religiosity), rather than from relationship behaviors per se (Collins and Madsen, 2006).
Reliance on Self-Report Data
Virtually all research on adolescent romantic relationships and development relies exclusively on self-report data. Thus, most of what is known about adolescents’ experiences and outcomes comes from their own descriptions. As has been widely discussed in the survey literature (e.g., Schwarz, 1999; Singer and Presser, 1989), self-reports can provide a limited or even distorted picture of a phenomenon, for several reasons. First, self-reports are subject to distortions because of failures of recall (i.e., individuals do not always remember their experiences accurately and so resort to educated guesses when responding to surveys) and deliberate bias (e.g., respondents may alter their answers to create a favorable impression for the researcher). Second, when an independent variable (e.g., relationship experiences) and a dependent variable (e.g., depression) are both assessed via self-reports, the association between them can be inflated by the fact that the measurements share a common method (e.g., Fincham and Bradbury, 1987). Third, there may be important aspects of adolescents’ lives that they simply do not have access to (e.g., their parents’ mental health history). There are several ways of broadening the range of research methods to overcome the limitations of self-report data, including collecting reports from multiple observers and collecting observational data, but these approaches have not been widely adopted.
Inadequate Controls for Exogenous Variables
Perhaps the most significant limitation of the existing literature on the long-term consequences of adolescent relationship experiences is the neglect of third-variable influences (Collins and Madsen, 2006; Halpern, 2003). Current interventions targeting the romantic relationships of adolescents assume not only that more successful adolescent relationships are associated with more successful adult relationships but also that experiences during adolescence are contributing causes of adult experiences. Evaluating the support for this assumption requires research that recognizes and carefully accounts for third-variable threats to causal validity (e.g., Shadish, Cook, and Campbell, 2002). The term third variable refers to a variable that has not been measured in a study but that may account for the association between two variables that have been measured. For example, early initiation of sexual intercourse may be associated with depression and behavior problems during adolescence (e.g., Aro and Taipale, 1987). One possible explanation for this association is that early initiation of sexual intercourse causes depression and behavior problems. However, it is also possible that the association between these two variables comes about only because they are both caused by an unmeasured third variable, such as socioeconomic status or level of parental supervision. The third variable therefore threatens causal inferences.
If associations between adolescent experiences and later young adult outcomes are products of third variables—or at least are in part due to such influences, whether genetic or environmental—then interventions that enhance adolescent relationships are much less certain to improve relationships in young adulthood. In contrast, if the associations between adolescent experiences and young adult outcomes can be observed even after controlling for genetic and environmental influences that affect both outcomes directly, then interventions that target adolescent relationships are more likely to achieve the desired improvements in adult outcomes. To the extent that research has ignored third-variable threats to causal validity, resulting estimates of adolescent experiences’ influence on later young adult outcomes may be inflated.
Conclusions
In general, research on the role of adolescent romantic relationships in adult development still awaits the rigor and methodological sophistication that could support strong conclusions. Lack of data does not appear to be the limiting factor in this area. For example, within such nationally representative, longitudinal data sets as Add Health and the NLSY lie data that could be used to examine central premises of current models. To date, however, definitive analyses of these data have not been conducted, leaving the existing research to rely primarily on studies of limited scope and generalizability. Keeping these limitations in mind, the following review evaluates the accumulated results of these studies, organized according to the specific paths suggested by the integrative framework presented in the previous chapter.
Antecedents of Adolescent Romantic Relationships
The integrative framework presented in Chapter Three suggests that adolescents’ early environments affect adult development through their direct effects on adolescent romantic relationships. Support for these models requires, as a preliminary step, evidence that adolescent experiences in romantic relationships are associated with their early environments. The emergent research on adolescent romantic relationships has examined these associations. Although, as noted above, much of this research relies on retrospective data, the results from cross-sectional and longitudinal studies of this question consistently point to strong associations between early environments and experiences and romantic relationships during adolescence.
Family of Origin
By far the most thoroughly studied antecedent of adolescent romantic relationships is the environment and structure of the adolescent’s family of origin. The best of these studies assess family variables during infancy and early childhood and then use those assessments to account for romantic experiences when those children have grown into adolescents and young adults. Research using this design has found evidence for substantial continuity between several distinct aspects of infants’ and children’s experiences in the family of origin and their subsequent experiences in romantic relationships.
Of particular interest to policymakers are the implications of family structure, i.e., whether a child is raised by two married parents. In this regard, research examining the intergenerational transmission of divorce has shown that children of divorced parents are significantly more likely to experience divorce themselves as adults (Amato, 1996; Amato and DeBoer, 2001). To explain this association, developmental psychologists, guided by attachment theory, have explored whether the structure of parents’ relationship is specifically associated with children’s orientations toward romantic relationship during adolescence. For example, Summers et al. (1998) compared the attachment styles of young adults whose parents either divorced or remained married during adolescence, finding that children of divorced parents reported less secure romantic attachments, even after controlling for demographic differences between the two groups. Others have noted that children of divorced parents express less optimism about their own marital prospects (e.g., Sprague and Kinney, 1997) and more negative attitudes toward marriage during adolescence (e.g., Tasker and Richards, 1994). Such findings are consistent with models suggesting that parental divorce influences adult marital outcomes through its direct effects on adolescents’ personal ideas and beliefs about relationships. However, as others have noted (e.g., Amato, 2001), there are several other, noncognitive paths through which parental divorce may affect the development of romantic relationships across the life course, including the cumulative effects of the reduced socioeconomic status and educational attainment that follow divorce (Tasker and Richards, 1994).
A relatively well-developed body of research has looked beyond family structure to examine how relationships among family members account for the development of romantic relationships in adolescents and adults. One especially rigorous demonstration of these associations involved recording interactions among family members in 99 families with adolescents and coding the quality of those relationships (Bell and Bell, 2005). The quality of the interactions accounted for significant variability in the well-being of the adolescents 25 years later. Similar research drawing from the British National Child Development Study also finds that good relationships with parents and siblings at age 16 is associated with more satisfying marriages at age 33 (Flouri and Buchanan, 2002). In other words, consistent with the integrative framework presented in Chapter Three, there is evidence of direct associations between relationships with parents during adolescence and well-being during adulthood.
Consistent with the idea that these associations operate through their direct effects on romantic relationships during adolescence, several studies have shown direct links between the quality of family relationships during infancy and early childhood and the quality of romantic relationships during adolescence and young adulthood. Some of this research has drawn upon cross-sectional and retrospective data that are subject to questions of bias and shared method variance (e.g., Benson et al., 1993; Feeney, Noller, and Patty, 1993; Furman et al., 2002; Murry et al., 2006; Reese-Weber and Marchand, 2002). The best of this research, however, draws upon longitudinal studies of children and their families over time, and has often been guided, explicitly or implicitly, by an attachment theory perspective. For example, the Minnesota Longitudinal Study of Parents and Children has followed the children of 267 first-time mothers from 1975 to the present and includes self-report and observational assessments of mother-infant interactions during infancy, as well as assessments of dating experiences during adolescence and early adulthood. Analyses of the multiple waves of data available for each child reveal significant continuity in the nature of the interpersonal relationships that these children experienced across stages of development (e.g., Carlson, Sroufe, and Egeland, 2004; Sroufe et al., 2005). Specifically, children whose interactions with their mother were coded as more positive during infancy grew up to have more secure representations of peer and romantic relationships during adolescence and subsequently had more successful romantic relationships in young adulthood (see also Collins et al., 1997; Seiffge-Krenke, Shulman, and Kressinger, 2001). Research focused more specifically on attachment styles finds similar links between the observed interaction between mothers and their infants and the attachment representations of those infants up to 20 years later (Roisman et al., 2005; Waters et al., 2000).
Other longitudinal research in this vein assesses children’s relationships with their parents during adolescence as predictors of their subsequent romantic experiences during young adulthood. For example, further analyses of data from the Minnesota Longitudinal Study of Parents and Children examined associations between videotaped interactions between parents and their 13-year-old children and those children’s subsequent interactions with their romantic partners at ages 20–21 (Roisman et al., 2001). Results revealed that children who interacted more positively with their parents at age 13 also interacted more positively with their romantic partners at ages 20–21. Moreover, this association appeared to operate through attachment styles in later adolescence (age 19). In other words, adolescent interactions with parents predicted later models of relationships in adolescents, and those models in turn accounted for the way adolescents treated their partners as young adults.
Conger et al. (2000), in an independent eight-year longitudinal study that also drew upon observational data, showed that when adolescents’ interactions with their parents were more involved and supportive, their later interactions with romantic partners were coded as more supportive and less hostile. These behaviors in turn accounted for the associations between relationships with parents and the quality of romantic relationships in young adulthood (Conger et al., 2000). In contrast, when higher levels of negativity were observed in parent-child interactions during adolescence, the later social relationships of those adolescents appear to suffer (Kim et al., 2001). A separate study that videotaped parent-child interactions when children were 13 years old found that children whose parents failed to respect boundaries were subsequently more vulnerable to experiencing intimate partner violence in young adulthood (Linder and Collins, 2005).
Although the general pattern of association between relationships with parents and relationships with romantic partners holds true for males and females, some studies have described gender differences in the specifics of these associations. For example, analyses of data from the Add Health study, while confirming that more positive feelings about parental relationships predict higher levels of satisfaction in romantic relationships seven years later, suggest that this association comes about through different mechanisms for males and females. For males, poor relationships with parents predict lower self-esteem, which in turn accounts for significant variance in their later romantic satisfaction. For females, self-esteem did not play a significant role in this relationship, but number of sexual partners did, so that girls reporting worse relationships with their parents had more partners, which helped to account for their lower romantic satisfaction in young adulthood (Joyner and Campa, 2006). An independent six-year study that followed 122 adolescents through early adulthood found stronger, but similar, associations between family variables and later romantic relationship variables for females than for males (Feldman, Gowen, and Fisher, 1998).
The general pattern of association between the early family environment and relationships with romantic partners is further confirmed by studies that have examined the implications of specific experiences within the family of origin. One that has received particular attention is the experience of physical or sexual abuse in childhood, both of which, not surprisingly, have been associated with less secure attachment styles in adolescence (Flanagan and Furman, 2000) and higher likelihood of experiencing and perpetrating aggression in later romantic relationships (Linder and Collins, 2005). Several studies further suggest that exposure to marital distress between parents also increases the child’s risk of experiencing dating violence and poorer relationship outcomes in adolescence and adulthood (e.g., Amato and Booth, 2001; Capaldi and Clark, 1998; Kinsfogel and Grych, 2004; Simons, Lin, and Gordon, 1998; Stein-berg, Davila, and Fincham, 2006). Finally, longitudinal data from the NSFH have been used to show that children whose parents monitor their activities prior to adolescence initiate dating and sexual behavior later during adolescence, compared with children whose activities were less closely monitored by their parents (Longmore, Manning, and Giordano, 2001).
It is hard to judge how well any of this work applies to low-income youth specifically. To the extent that instability and distress in the family of origin predict worse outcomes in later romantic relationships, children in low-income populations should be at greater risk for experiencing relationship problems, because these populations experience higher levels of family instability and distress. A rare study that examined these issues in low-income populations examined cross-sectional and retrospective data from a sample of 267 black female adolescents from impoverished neighborhoods (Adam and Chase-Lansdale, 2002). Among these young girls, reports of having been separated from their parents and reports of higher numbers of residential moves were associated with higher levels of adjustment problems, including sexual behavior problems, even after adjusting for demographic and economic variables. Although these are not the same variables that have been examined in general population samples and white samples, the results do paint a similar picture, suggesting that early experiences of instability in the family of origin have lasting consequences for the development of healthy relationships.
In sum, within a literature that is still in its early stages of development, no question has been studied more extensively than the implications of aspects of the family of origin. It is promising that, despite a wide range of operationalizations and methodologies, the results of this work are generally consistent across studies. When the early family environment of the child is relatively stable and when parents are warm and attentive, the child is more likely to experience more satisfying romantic relationships in adolescence and beyond. When the early family environment is characterized by distress or disruption, the child is at greater risk for experiencing relationship problems.
With these associations established, how these associations come about remains to be explored. Research guided by attachment theory finds evidence of cognitive mechanisms, such that experiences with parents shape the ideas and beliefs about relationships that young people use to guide choices in their own romantic relationships. Behaviorally oriented work finds evidence of skills-based mechanisms, such that interactions with parents form the basis for subsequent interactions with romantic partners. Both lines of work imply that, absent interventions, development through adolescence and into adulthood should be characterized by continuity, such that adolescents with maladaptive patterns of belief and behavior grow into young adults with the same maladaptive patterns. These findings therefore support the idea of cognitive and behavioral interventions during adolescence to sever these links and offer alternative models of thinking and behaving in romantic relationships to adolescents who might otherwise be at risk.
However, such explanations have yet to be pitted against third-variable explanations for these associations, such as genetic and contextual influences that may affect both family relationships and romantic relationships directly. Until alternative hypotheses have been ruled out definitively, causal statements about the effects of the family of origin on healthy adult relationships must be considered premature.
Individual Characteristics
In addition to the family of origin, enduring characteristics of the child have also been thought to account for continuities in relationships across the lifespan. For example, to the extent that personality is relatively stable over time, personality observed in infancy and early childhood may be associated with relationship outcomes in adolescence and young adulthood.
Caspi and his colleagues have been leaders in exploring this possibility, drawing upon studies that follow individuals across their lives to demonstrate the implications of childhood personality for adolescent and adult relationships. In a series of studies drawing from the Berkeley Longitudinal Study, Caspi and colleagues identified the paths through which different temperaments observed in early childhood gave rise to specific patterns of outcomes in later life (Caspi, 1987; Caspi, Bem, and Elder, 1989; Caspi, Elder, and Bem, 1987, 1988). Children judged by parents and teachers as explosive during childhood experienced higher rates of divorce as adults, and these associations were mediated by the cumulative educational, economic, and social consequences of childhood personality (Caspi, Elder, and Bem, 1987).
A more recent longitudinal study by these researchers confirmed these effects, demonstrating in a sample of over 900 children that undercontrolled temperament observed at age 3 predicts greater levels of conflict in romantic relationships at age 21 (Newman et al., 1997). Cross-sectional research on children self-identified as bullies suggests that such children experience deficits in their romantic relationships during adolescence specifically (Connolly et al., 2000). Research drawing upon the Minnesota Longitudinal Study refines this picture, exploiting the multiple assessments of this study to show how sociability and impulsivity assessed at less than 3 years of age predicts earlier initiation of romantic relationships in early adolescence, which in turn predicted greater alcohol use and a higher number of sexual partners by age 19 (Zimmer-Gembeck, Siebenbruner, and Collins, 2004).
Together, these studies suggest that an important way that personality affects adult relationship and marital outcomes may be through its direct effects on the timing of entry into dating and sexuality during adolescence.
A second individual characteristic that may also affect the initiation of romantic relationships is the timing of puberty and sexual maturation. Independent of personality or social context, some children enter puberty earlier or later than others. Most research on the effects of pubertal timing in adolescence has focused on its association with the initiation of sexual activity (Halpern, 2003). To date, the accumulated research on this effect indicates that earlier entry into puberty, however it is measured, predicts earlier sexual activity. For example, in the Add Health study, “early maturers” are more likely to have first sexual intercourse at a younger age compared with girls who mature later, particularly white early maturers (22 percent of white early maturers compared with 10 percent of white later maturers, p < 0.01) (Cavanagh, 2004). Girls who mature earlier also have fewer same-sex friendships and more boys in their friendship groups compared with other girls (Cavanagh, 2004). Although this finding has been consistent across several studies, Halpern (2003) points out that pubertal timing generally accounts for a relatively small amount (between 3 percent and 5 percent) of variance. Nevertheless, for girls, early physical maturation appears to bring about later negative outcomes like substance abuse and poor academic achievement through its direct association with sexual behavior (Caspi et al., 1993; Stattin and Magnusson, 1990).
For boys, in contrast, at least one longitudinal study finds evidence that pubertal timing may have positive long-term implications for life outcomes. Data from the Terman study, a longitudinal project begun in the 1920s that assessed gifted children repeatedly across their lives, revealed that boys who entered puberty earlier went on to experience higher levels of career and marital success (Taga, Markey, and Friedman, 2006). This is a provocative finding, suggesting biological influences on lifespan development, but its implications for understanding the impact of pubertal timing on less gifted youth in the present day are probably limited. For example, more recent analyses of data from the Add Health study finds early pubertal timing in boys to be associated with greater risk of being in physical fights, getting shot, or getting stabbed (Haynie and Piquero, 2006).
Research that directly examines associations between pubertal timing and dating behaviors, as opposed to sexual behaviors specifically, has found only modest effects. For example, analyses of data from adolescents 12–17 years old from the U.S. National Health Examination Survey examined whether levels of sexual maturation accounted for dating behaviors after controlling for the simple effects of age (Dornbusch et al., 1981). Whereas age was significantly associated with dating, so that older adolescents reported more dating, levels of sexual maturation accounted for little additional variance, suggesting that dating behaviors, more than sexual behaviors, may be shaped by social pressures rather than biological factors.
To the extent that temperament and pubertal timing may play a role in the development of adolescent romantic relationships, there may be genetic sources of influence on romantic relationships. To date, no research has examined genetic influence on relationship behaviors during adolescence, but behavioral genetic studies of identical and fraternal twins have demonstrated substantial heritabilities for divorce in adulthood (Jockin, McGue, and Lykken, 1996; McGue and Lykken, 1992). These genetic effects may be mediated by relationship behaviors during adolescence, but the intervening steps in this association have not been explored.
None of the research on individual characteristics in relationship development has addressed low-income populations specifically. In the absence of data, it is possible to speculate that the effects of individual characteristics on later romantic relationships are likely to be exacerbated by an impoverished environment. Exploring a similar idea, Caspi and his colleagues (Caspi et al., 1993) showed that the impact of early menarche on delinquency in adolescent girls is moderated by the sex composition of the schools they attend, such that early menarche predicts greater risk for girls at coeducational schools than at single-sex schools. It seems likely that similar interactions occur between individual characteristics and socioeconomic status, so that individual characteristics that place adolescents at risk in the general population have even greater effects in low-income contexts. Research confirming such a hypothesis could be used to target adolescents who are especially likely to be vulnerable and offer them specific or additional interventions.
Correlates of Adolescent Romantic Relationships
The distinction between antecedents and correlates of adolescent romantic relationships is more theoretical than empirical. In theory, the antecedent variables described in the previous section are in place long before the adolescent enters a romantic relationship for the first time. Because they precede the experience, antecedent variables are potential causes of relationship outcomes. The correlates described in this section, in contrast, are more proximal to the relationship experiences. They may precede relationship experiences, but they may also co-occur or change along with those experiences. In practice, because much of this research is crosssectional, the causal relationships between adolescent relationship outcomes and the variables described in both sections remain an open question.
Peer Groups
Developmental models of adolescent relationships suggest that relationships with parents affect relationships with romantic partners in part through their direct effects on friendships with peers (e.g., Collins, 2003; Furman and Wehner, 1997). Perhaps as a result, peers are second only to the family of origin in the frequency with which they have been studied in research on adolescent romantic relationships. Stage models of the development of romantic relationships in adolescence propose an orderly sequence that begins with relationships with parents in infancy predicting the development and quality of relationships with same-sex friends in childhood. As the child enters adolescence and these early peer groups develop, they grow to include opposite-sex friendships, which in turn, for most adolescents, set the stage for the initiation of romantic relationships. Thus, peer networks form a proximal context for the initiation of romantic relationships in adolescence, and so should be associated with the timing and the quality of those relationships (Connolly and Goldberg, 1999; Furman, 1999).
Longitudinal research on relationships from childhood through late adolescence supports the general sequence of relationship experiences that stage models have proposed. In a three-year study of 180 high school students, having small groups of friends in childhood predicted having opposite-sex friends in early adolescence, which in turn predicted the initiation of romantic relationships in later adolescence (Connolly et al., 2000). A second study of an ethnically diverse sample of 1,284 adolescents observed the same progression (Connolly et al., 2004). To date, there have been no published attempts to determine whether this sequence replicates within low-income adolescents, or within specific racial or ethnic groups.
The majority of research on associations between peer and romantic relationships during adolescence has sought to move beyond the sequencing of these relationships toward an understanding of the aspects of peer relationships that contribute to more or less favorable romantic outcomes. To this end, a number of studies have examined how the size of an adolescent’s peer network is associated with the timing and quality of his or her romantic relationships, with somewhat mixed results. For example, one longitudinal study of 92 white, middle class adolescents tracked the development of their relationships when they were 9, 13, and 18 years old. For both boys and girls, having a larger peer network in late childhood predicted having closer romantic relationships in middle adolescence, and in turn more enduring romantic relationships in later adolescence (Feiring, 1999b). The implication of these results is that having more friends in early adolescence is associated with qualities that promote healthier relationships. A separate study found that high school students rated as more popular by their peers also report dating more frequently (Franzoi, Davis, and Vasquez-Suson, 1994).
However, analyses of data from the Minnesota Longitudinal Study of Parents and Children paint a more complicated picture. Drawing upon the multiple assessments of that study, Zimmer-Gembeck, Siebenbruner, and Collins (2004) found that sociability and peer acceptance in early adolescence predicted both earlier entry into romantic relationships and greater use of alcohol. Together, these variables in turn predicted a greater number of sexual partners by the age of 19, not the result that would be expected based on the Feiring (1999b) study. The fact that the 2004 study explicitly measured alcohol use may help to account for the difference. Whereas having friends proves to be a necessary first step toward the development of sexual and romantic relationships in both studies, a large social network that includes alcohol might mean something very different from a social network of similar size that does not regularly use alcohol. Although these issues have yet to be examined in low-income populations specifically, these findings suggest that in populations where alcohol and other drugs are readily available, larger networks of friends may not benefit the development of healthy romantic relationships.
In addition to the size of the peer network, characteristics of peer and friendship groups have also been associated with age at which adolescents enter romantic and sexual relationships. In the NLSY study, children of the 1979 cohort who at ages 11–12 had primarily same-sex friendships and whose friendships tended to be with preteens in the same grade were significantly less likely to report weekly dating at ages 13–14 and 15–16 (Cooksey et al., 2002). Younger adolescents whose friends were primarily in higher grades were significantly more likely to become sexually experienced between ages 13–14 and 15–16. A study using Add Health data found a direct association between having friends who are sexually experienced and an adolescent becoming sexually experienced as well. As the authors noted, “for every 1 percent increase in sexually experienced friends at Wave 1, the odds that young people initiated sex by Wave 2 increased by 1%” (Sieving et al., 2006). In other words, in terms of shaping adolescents’ entry into romantic relationships, what matters is not just the size of the peer network but also the composition of that network.
The associations between the quality of relationships with peers and with romantic partners are much clearer. Across several independent samples, adolescents’ reports of the quality of their peer relationships are significantly correlated with their reports of the quality of their romantic relationships, such that adolescents who describe their friendships as close and supportive tend to describe their current and subsequent romantic relationships in similar terms (e.g., Collins et al., 1997; Connolly et al., 2000; Furman et al., 2002; Seiffge-Krenke, Shulman, and Kiessinger, 2001). Problems within peer groups may also have consequences for romantic relationships in adolescence. At least among white female adolescents, interpersonal problems in peer groups are associated with greater likelihood of initiating sexual intercourse (Cavanagh, 2004). Supporting the idea that peer relationships presage a new experience of intimacy for adolescents, two studies have explicitly shown that ratings of adolescent romantic relationships are more strongly correlated with ratings of peer relationships than with ratings of relationships with parents (Furman et al., 2002).
In addition to providing opportunities for developing romantic relationships, peers also seem likely to shape the standards that adolescents apply toward those relationships (Brown, 1999). Early research on peer relationships and sexual behavior found that adolescents tend to have friends whose levels of sexual experience match their own (Billy and Udry, 1985). Anticipated peer approval or disapproval is also a predictor of intercourse initiation among youth (Kirby, Lepore, and Ryan, 2005). Among gay and bisexual youth, the perception that certain factions in the gay community disapprove of romantic relationships may discourage participation in them (Myer, 1989). Such results are consistent with the idea that friends affect each other’s decisions about when to enter romantic relationships and engage in sex, although they may also be evidence of a similarity bias in selecting friends.
Research on physical aggression in adolescent relationships offers additional evidence of a peer influence on standards within relationships. In an analysis of data from 391 adolescents, boys who believed that aggression was more common in the relationships of their peers engaged in more aggression in their own relationships (Kinsfogel and Grych, 2004). These data were collected at a single assessment, so it is not possible to infer a causal relationship from this association, but the result is consistent with the idea that peers teach each other the sorts of behaviors that are and are not appropriate within romantic relationships. It is also possible, however, that adolescents choose peers whose standards match their own, leaving the role of peers in setting standards for romantic relationships an open question for now.
Although it is common for peer relationships to be described as potential causes of subsequent romantic relationships, at least one study has described how the experience of romantic relationships may affect developing peer relationships during adolescence. Drawing upon retrospective data from 102 white female adolescents, Zimmer-Gembeck (1999) found that spending more time with a romantic partner was associated with spending less time with friends.
In general, research on the potential influence of peers on the development of romantic relationships in adolescence has been more suggestive than definitive. The idea that peer relationships form a crucial stepping stone from the intimacy of the family of origin to the intimacy of romantic relationships has received consistent support, but exclusively from studies of a relatively narrow range of adolescents. More precise questions about the paths through which relationships with peers may shape subsequent relationships with romantic partners have only been touched upon, and again never in research on low-income youth specifically. Although evidence from cross-sectional and longitudinal research is consistent with current theories about how peers may affect the development of romantic relationships, the causal role of peers in affecting the standards that adolescents apply to their relationships and the quality of those relationships has yet to be determined.
Qualities of the Relationship
Within adult relationships and marriages, a number of studies have explored which specific elements of the relationship are most strongly associated with partners’ satisfaction. In adolescent relationships, several researchers have taken up this issue, but to date these studies have relied almost exclusively on cross-sectional self-report data from convenience samples, and so may be of limited value. For example, one study that examined self-report data from 304 adolescents found correlates of relationship satisfaction that generally echoed the correlates of satisfaction in adult relationships, e.g., commitment, being appreciated, good communication, feeling attracted to the partner (Levesque, 1993).
Another set of studies on this issue draws on assessments of 61 adolescent couples to examine correlates of relationship satisfaction and individual well-being (e.g., Galliher et al., 1999). Cross-sectionally, male and female partners who were more satisfied with their relationships tended also to be more satisfied with their communication (Galliher et al., 2004) and their relationships were less likely to end during the subsequent year (Rostosky et al., 2000). In a separate analysis of these data, these researchers also examined associations between feelings of commitment and sexual behavior, finding, perhaps counterintuitively, that physical expressions of affection (e.g., holding hands, kissing) were more strongly correlated with feelings of commitment than was engaging in sexual intercourse (Rostosky et al., 1999). Were this finding to be replicated in a larger and more representative sample, it would suggest that, at least for some adolescents, expressing physical affection reflects deeper commitment to a relationship than does engaging in sex.
A number of other qualities of adolescent romantic relationships may be associated with the concurrent and longitudinal implications of those relationships. For example, the Add Health study has used an innovative card sorting technique to assess the sequencing of specific behaviors within adolescent relationships (i.e., whether couples say “I love you” prior to or after having sex, when partners meet each other’s parents, when partners appear in public as a couple). The timing of these behaviors may help to determine the impact of a specific relationship on the developing adolescent; to date, however, data from this part of the Add Health study have not been examined in published research. Other potential relationship qualities of interest might include the amount of time that adolescent partners spend together, how couples spend that time, and the specific age within adolescence when these relationships occur. None of these issues has yet been addressed in this area. Moreover, although it seems likely that the correlates of relationship satisfaction may vary significantly across low-income and upper-income populations, this question has yet to be addressed empirically.
Partner Characteristics
As noted earlier, virtually nothing is known about the ways that characteristics of adolescents’ romantic partners affect the implications of adolescent romantic experiences (Collins and Van Dulmen, 2006). This seems a serious oversight, especially given Rutter’s early research showing that characteristics of a partner can make a significant difference to the adult outcomes even of children who come from severely disadvantaged backgrounds (Rutter and Quinton, 1984).
Two studies that have addressed this issue examined the implications of a partner’s age. For example, a study of 2,829 students recruited from sixth-grade classes in 19 urban schools assessed the presence of a partner, partner age, and sexual behaviors (VanOss Marín et al., 2000). Compared with those without steady partners, students with partners more than two years older than themselves were more likely to be Hispanic, less likely to be acculturated, and over 30 times more likely to have had sex. In research on a middle school intervention to reduce sexual risk behaviors, the association between relationship status and sexual behavior was examined in detail for 2,829 seventh-, eighth-, and ninth-grade students in 19 predominantly Hispanic schools in urban Northern California (VanOss Marín et al., 2006). Having had a boyfriend or girlfriend in seventh grade increased the chances that the student had been sexually active by ninth grade. Having had a relationship with someone their own age doubled their odds of being sexually active in ninth grade, and girls who had a relationship with a boy in a higher grade by seventh grade were also more likely to be sexually experienced by ninth grade.
A smaller study of 146 girls recruited from ninth-grade classes found similar results, showing that girls whose partners were more than three years older than themselves were more likely to be engaging in all forms of sexual activity, more likely to be having sex while using drugs or alcohol, and more likely to report having been coerced into having sex, compared with girls whose partners were closer to their own age (Gowen et al., 2004). In terms of their attitudes and beliefs, girls with older boyfriends were more likely to endorse the belief that engaging in sex is associated with maturity. The authors of this study viewed these differences as evidence of power imbalances in the relationships of girls with older boyfriends, although power was not assessed directly.
To the extent that power imbalance is a salient issue in adolescent romantic relationships, it is possible to imagine numerous other partner characteristics that may lead to such imbalances, including physical attractiveness, physical strength, wealth, race and ethnicity, and social status. The associations among these partner characteristics and experiences in adolescent romantic relationships have not been studied.
Beliefs and Attitudes
Considering the number of studies that have described the beliefs and attitudes of adolescents with regard to romantic relationships and marriage (reviewed in Chapter Two), it is perhaps surprising that so few studies have examined whether adolescents’ beliefs and attitudes are in fact associated with their experiences and behaviors in romantic relationships. Most studies that have examined links between cognitive variables and adolescent behaviors have focused narrowly on sexuality rather than romantic relationships.
For example, attitudes toward sexuality and exposure to sexuality in the media have been associated with sexual intentions and behaviors in several studies of adolescents (e.g., Collins et al., 2004; Pardun, L’Engle, and Brown, 2005). Analyses of the 1995 NSFG indicate that religious affiliation and frequency of attendance at religious services, variables that may be proxies for more conservative attitudes toward sexuality, account for the timing of first intercourse among adolescents, such that stronger affiliation and more frequent attendance are associated with delays, even after controlling for demographic variables (Jones, Darroch, and Singh, 2005). Intelligence, as measured by scores on the Peabody Picture Vocabulary Test (PPVT), is also associated with delays in initiating sex, according to analyses of data from Add Health (Halpern et al., 2000).
Yet aside from sexual behaviors, little is known about how other behaviors and choices relevant to romantic relationships are associated with adolescents’ beliefs and attitudes. One exception is a cross-sectional study of 254 secondary school students that examined attitudes toward aggressive behavior in romantic relationships (Feiring et al., 2002). Although aggressive behaviors were rare in this sample, males and females who reported having engaged in aggressive behavior toward a romantic partner also endorsed statements indicating that aggressive behavior toward a romantic partner could be justified.
Links between other sorts of attitudes (e.g., beliefs about the importance of fidelity, the role of sexuality in intimacy, appropriate demonstrations of affection, the importance of relationships compared with other activities) and corresponding behaviors are unknown. In addition, no research linking attitudes to behaviors has been conducted among low-income adolescents. Given the prominent role that beliefs and attitudes play in the interventions currently being targeted toward adolescents in this population, such a lack of data must be considered a serious oversight. Filling this gap should be an immediate priority for research in this area.
Immediate Consequences of Adolescent Romantic Relationships for Adolescents
Accumulating evidence suggests that adolescents’ experiences in romantic relationships have immediate and significant consequences for their well-being during adolescence. What is not currently clear is the extent to which experiences in romantic relationships may be said to cause those consequences. The studies reviewed in this section frequently demonstrate associations between specific experiences in relationships and specific outcomes during adolescence. Supporting a causal link between the experience and the outcome requires ruling out potential third variables that may in fact cause both the experience and the outcome. In cross-sectional studies, the problems are compounded by uncertainty about whether variables treated as outcomes in fact preceded the relationship experiences. The existing research in this area has rarely ruled out these alternative hypotheses. Thus, the studies described here can be said to have identified experiences associated with certain outcomes, but whether the experiences cause the outcomes (and so might appropriate targets of interventions designed to improve outcomes) remains an open question.
Psychological Well-Being
In a frequently cited paper drawing on data from the Add Health study, Joyner and Udry (2000) found that, relative to adolescents not involved in romantic relationships, adolescents in romantic relationships experienced greater increases in depressive symptoms over the period of one year, and this increase was greater for females than for males. An earlier review made a similar point, suggesting that early entry into dating is associated with lower self-esteem (Simmons et al., 1979). Girls who are involved in romantic relationships are also at higher risk for eating disorders (Cauffman and Steinberg, 1996; Smolak, Levine, and Gralen, 1993).
What may account for these associations? To determine whether the initiation of dating places youth at risk, researchers working with the Add Health data identified 2,344 participants in the initial assessment who had never had sex and never dated and examined their experiences and symptoms over the subsequent year (Grello et al., 2003). Adolescents who began dating, and those who had their first sexual experience in the context of a romantic relationship, did not experience significant changes in their depressive symptoms or delinquent behaviors. However, engaging in casual sex was associated with higher levels of depressive symptoms and problem behaviors.
Studies that have looked beyond relationship status to address qualities and processes in adolescent relationships offer a more refined picture. For example, research that assessed the degree of closeness in the relationships of 72 girls in late adolescence found that a lack of closeness in romantic relationships, but not a lack in other kinds of relationships, was associated with stronger tendency to experience negative moods (Williams, Connolly, and Segal, 2001). In a sample of 421 adolescents, those who described negative qualities in their romantic relationships also reported higher levels of depressive symptoms (La Greca and Harrison, 2005). Both these studies suggest that it is the quality of the romantic relationship, rather than the relationship status per se, that is associated with depressive symptoms during adolescence.
Consistent with this view, data from an epidemiological sample of 1,470 adolescents reveal that experiencing the breakup of a romantic relationship is a significant predictor of the first onset of major depression (Monroe et al., 1999). In contrast, a study of 480 Israeli adolescents found that those in steady dating relationships reported higher self-esteem and were rated more positively by their peers as well (Samet and Kelly, 1987).
Such findings place the Joyner and Udry (2000) research in perspective: Romantic relationships during adolescence provide an opportunity for emotional pain that puts young people at risk for experiencing depression, but this risk may be linked to unhealthy or unsuccessful relationships (Welsh, Grello and Harper, 2003). Thus, the converse may also be true: Healthy relationships may offer opportunities for growth and fulfillment that improve well-being and increase resilience. In other words, the quality of the romantic relationships that adolescents experience may moderate the implications of those experiences for their psychological well-being, just as is true in adult marriages (e.g., Beach, 2001). To date, there has been no research on these issues in low-income populations, but there is no reason to believe that the role of relationship quality is any less important in that population than in more affluent populations.
Sexual Behavior and Contraception
Experiencing a romantic relationship within the past 18 months is one of the most powerful predictors of sexual activity among adolescents (Blum, Beuhring, and Rinehart, 2000; Halpern et al., 2000; Miller et al., 1997), especially among girls (Marin et al., 2006). Because a romantic relationship provides a context that makes sex possible and rewarding, those who enter relationships earlier also tend to engage in sexual activities earlier ( Thornton, 1990). In a sample of virgins ages 15–21 drawn from Add Health, the number of romantic relationships in the prior 18 months was strongly associated with likelihood of first intercourse in the subsequent year (Rostosky, Regenerus, and Wright, 2003). Analyses of data from the NLSY suggest that the romantic relationship per se, rather than mere dating, predicts sexual behavior: Having a steady partner accounts for first experiences of sex, but frequency of dating does not (Cooksey, Mott, and Neubauer, 2002). Research on a small sample of Australian young people ages 16 to 25 similarly found that, as dating relationships became more serious and committed, the desire for sexual intimacy and actual sexual behavior both increase (McCabe and Collins, 1984). Indeed, the modal reported reason for first intercourse among teens is to have the partner love them more (Rodgers, 1996). It is unknown whether these associations are replicated among low-income populations specifically.
Given the strong associations between romantic relationships and sexual behavior, an important question for policymakers is whether romantic relationships make safe sex and the use of contraception more or less likely among adolescents who are sexually active. Unfortunately, the existing research on this issue has produced mixed results (Giordano, 2003). Some studies suggest that involvement in romantic relationships is associated with more consistent use of condoms. For example, data from a subsample of adolescent girls in the 1995 NSFG survey who had had sexual intercourse for the first time before age 18 showed that roughly half (52 percent) of the female adolescents who had just met their sexual partner prior to having sexual intercourse used no method of contraception, compared with 24 percent of girls who reported that they were “going steady” with their partner. This relationship remained statistically significant when other variables linked to the likelihood of contraceptive use were controlled (Manning, Longmore, and Giordano, 2000). Analyses of data from the Add Health study also indicate that female adolescents who were acquainted with their partners before they were involved in a romantic relationship were more likely to use birth control consistently, compared with females who did not know their partners at all prior to entering the relationship (Kaestle et al., 2005) and that couples who are more similar to each other report more consistent condom use than couples who are less similar (Ford, Sohn, and Lepkowski, 2001). A smaller study of 625 inner-city black females, approximately one-third of whom were between ages 17 and 20, similarly observed that those who felt emotionally closer to their partners expressed stronger intentions to use condoms (Santelli et al., 1996). Together, these studies support the view that being in a romantic relationship—in particular, being in an emotionally close romantic relationship—is associated with higher levels of safe sex behavior among adolescents.
Yet several other studies have found no association between being in a relationship and condom use, or they have even found associations in the opposite direction. For example, a recently published study of 1,316 adolescents recruited from clinical settings found no difference in rates of self-reported condom use between those with a single partner and those who were sexually active with multiple partners (Lescano et al., 2006). Analyses of data from the NSAM indicate that, among sexually active adolescent males in exclusive romantic relationships, condom use declines as the length of the relationship increases (Ku, Sonenstein, and Pleck, 1994). Research on a smaller sample of 172 adolescent women paints a similar picture, showing that those in established relationships are quicker to have unprotected sex than those in new relationships (Fortenberry et al., 2002). By themselves, the results of these studies would suggest that adolescents in the strongest romantic relationships may be the least vigilant about contraception and safe sex.
How can we reconcile these two apparently contradictory sets of results? Clearly, more research is sorely needed to address this question. Still, in the absence of that research, we speculate that the explanation for these competing results may lie in the fact that, over the course of a single romantic relationship, adolescents’ use of condoms changes over time. For example, a recent longitudinal study interviewed 176 adolescent women every three months for two years, asking about relationship status and use of condoms in each interview (Sayegh et al., 2006). Among those in ongoing relationships, condom use declined significantly over time. Yet condom use started out at the highest level among those who were in the most satisfying relationships at the initial interview. Together these results suggest that the predictors of condom use early in a relationship may differ from the predictors later in the same relationship. Early in a relationship, when adolescents are initiating sexual activity for the first time, they may be more comfortable discussing and enacting safe sex behaviors with partners that they feel closer to and trust. Over time, however, deepening trust in a partner may, ironically, encourage adolescent partners in lasting relationships to be less vigilant about condom use than partners in newer relationships.
Ethnographic data from low-income populations offers some support for this sort of process. For example, interviews with adolescent black females reveal their concerns that the use of contraception indicates a lack of intimacy and trust for a partner (Dash, 2003). To the extent that such sentiments generalize to low-income adolescents more broadly, those in more serious romantic relationships would be expected to use condoms less frequently than those in less established relationships, even if those in more serious relationships were more likely to use condoms initially. Evaluating this possibility requires longitudinal research that assesses trajectories of trust and condom use over time in adolescent romantic relationships, but to date we are aware of no studies that have collected this sort of data.
Intimate Partner Violence
As noted in Chapter Two, physical aggression between intimate partners is an all-too-frequent occurrence within adolescent romantic relationships. Data from Wave I of the Add Health study, for example, reveals that 12 percent of respondents had been the victim of physical violence from a dating partner in the previous 18 months, and rates are similar for males and females (Halpern et al., 2001). Because this is an outcome of particular concern to researchers and policymakers, research on the causes and correlates of violence between partners in adolescent relationships is somewhat more extensive than research on other outcomes has been. In recent years, two excellent reviews of research in this area have been published (Capaldi and Gorman-Smith, 2003; Wekerle and Avgoustis, 2003), so the current review will only briefly touch on that work here.
The dominant theory of relationship violence among adolescents focuses on the idea of intergenerational transmission, i.e., the idea that antisocial behaviors in parents, including violence toward the child, leads to children who view aggression between intimates as acceptable, and in turn makes aggressive behavior toward romantic partners more likely (Capaldi and Gorman-Smith, 2003). A number of longitudinal studies have found support for this general progression for males and females (e.g., Andrews et al., 2000; Capaldi and Clark, 1998; Magdol et al., 1998; Simons, Lin, and Gordon, 1998). Analyses of data from Add Health further suggest that violence is more likely to occur between adolescent romantic partners who have had sex than between partners who have not had sex (Kaestle and Halpern, 2005b). Additional analyses of the Add Health data reveal that, for both genders, engaging in intimate partner violence is significantly associated with increasing age and having a higher number of intimate partners (Roberts, Auinger, and Klein, 2005). Across a number of domains (e.g., socioeconomic resources, family relations, educational achievements, and problem behaviors), variables assessed during adolescence predict partner abuse at 21 more effectively than variables assessed prior to adolescence, suggesting that adolescence may be a crucial period during which the antecedents of partner violence take shape (Magdol et al., 1998).
Two longitudinal data sets have been particularly useful in examining the antecedents and correlates of partner violence in adolescence. One, the Oregon Youth Study, comprises primarily white males who were first assessed in fourth grade and have been followed yearly thereafter (Capaldi and Patterson, 1987). A second, the Chicago Youth Development Study, solicited fifth- and seventh-grade males from ethnically diverse communities with relatively high rates of poverty. Thus, the Chicago Youth Development Study provides specific insight into low-income youth, in marked contrast to most of the research reviewed in this report. It is especially noteworthy that the results of tests of the intergenerational model obtained from the Oregon Youth Study have generally replicated in the Chicago Youth Development Study, suggesting that the antecedents of partner violence may be similar across low-income and more affluent populations (Capaldi and Gorman-Smith, 2003). Both studies provide results supporting the idea of intergenerational transmission, showing that antisocial behaviors among parents predicts aggressive behaviors in their adolescent children (e.g., Capaldi and Clark, 1998).
One reason that understanding partner violence is particularly important for the current report is that the experience of partner violence during adolescence seems likely to have long-term implications for an individual’s orientation toward future romantic relationships. The existing research in this area has yet to address these implications directly. There has been some research showing continuity among perpetrators of relationship violence, such that adolescents who are more accepting of violence against a partner are more likely to engage in violence within relationships as young adults (e.g., Capaldi et al., 2001). In addition, among female adolescents in the Add Health study, abuse by an intimate partner predicted significant increases in illicit substance use, antisocial behavior, and suicidal behavior one year later (Roberts, Klein, and Fisher, 2003). Studies of victims of partner violence during adolescence have not yet examined their consequences for outcomes assessed in adulthood.
Substance Abuse and Academic Achievement
Although substance abuse and academic achievement are quite different outcomes, they have often been examined in the same studies, perhaps because they tend to be negatively correlated (i.e., engaging in substance abuse is strongly associated with lower levels of academic achievement). Much of the research on the implications of romantic relationships for these outcomes has focused on the effects of early or later entry into romantic relationships on substance abuse and academic achievement in girls. The results of this research have been consistent across multiple studies: The earlier that girls become involved in romantic relationships, the higher their risks of later substance abuse and the lower their academic achievement (e.g., Aro and Taipale, 1987; Grinder, 1966; Pawlby, Mills, and Quinton, 1997). In later adolescence, the negative implications of romantic relationships appears to subside (Neemann, Hubbard, and Masten, 1995), but early entry into marriage continues to be associated with increased risk of substance abuse (e.g., Bachman et al., 1997; Schulenberg et al., 2005).
Throughout these studies, researchers have occasionally succumbed to the temptation to infer causal links from correlational data, suggesting that early involvement in romantic relationships causes later problems for girls. In fact, the nature of the causal influence in these studies is far from clear, because behavior problems that may predict later substance abuse and academic deficits have also been shown to predict early entry into romantic relationships (e.g., Aro and Taipale, 1987; Connolly et al., 2000). Nor is it clear that the same patterns of association would hold true for males. Indeed, Giordano (2003) speculates that early entry into romantic relationships may be positive for adolescent boys, although no data have addressed this possibility.
To the extent that early entry into romantic relationships is associated with a higher risk of substance abuse and lower academic achievement, at least for girls, late entry into romantic relationships may be associated with positive outcomes. For example, data from the Add Health study have been used to show that adolescents who postpone sexual activities score higher on measures of intelligence (Halpern et al., 2000). Other analyses of Add Health find that, compared with adolescents in dating relationships, those who do not date have the highest grades and the lowest levels of delinquent behavior (Giordano, Manning, and Longmore, 2006). Again, given the current state of the research on these issues, these associations are as likely to be the result of simple selection effects (i.e., students with greater academic prospects do not choose to devote time to romantic pursuits) as they are to reflect any causal effects of romantic relationships on academic outcomes.
The true relationship between romantic relationships and such outcomes as substance abuse and academic achievement is likely to be more complex than a simple focus on relationship status would suggest. For example, analyses of data from the Toledo Adolescent Relationships Study reveal that respondents’ grades were significantly predicted by the grades of their romantic partners, controlling for parental monitoring (Giordano, Manning, and Longmore, 2006). It seems plausible that partners similarly affect substance abuse and delinquency. To the extent that partners do influence each other’s behaviors in this regard, the closeness of the relationship may moderate this influence, such that close partners influence each other more than partners who are less close. In other words, it may be that the characteristics of the relationship interacting with the characteristics of the partner, rather than the mere presence of a romantic relationship, account for associations between romantic experiences, substance abuse, and academic achievement in adolescents. In the absence of data, these hypotheses must remain speculative.
Consequences of Adolescent Romantic Relationships for Outcomes in Adulthood
Current policies that target adolescents for relationship education are motivated less by the desire to promote favorable outcomes during adolescence (although that is clearly one goal) than by the desire to promote healthy marriages in adulthood. Thus, the guiding assumptions of current policies are (1) that experiences with romantic relationships in adolescence are associated with subsequent marital outcomes in adulthood, and (2) that the nature of these associations is causal, such that intervening during adolescence is likely to lead to improved outcomes during adulthood. Evaluating these assumptions requires, at minimum, long-term longitudinal research that follows adolescents into adulthood, assessing the nature of their romantic relationships at each stage of development. As many scholars in this area have noted (e.g., Collins and Van Dulmen, 2006; Furman and Shaffer, 2003; Giordano, 2003; Kan and Cares, 2006), almost no research of this type has been published to date. Data capable of addressing these issues have been collected in the Add Health and NLSY studies, but published research has tended to draw upon samples that have not yet extended far into adulthood.
The research that has attempted to link adolescent relationships to adult marital outcomes has so far been extremely limited in scope. For example, research focused specifically on the consequences of teen pregnancy has identified the economic and educational deficits that teen mothers experience relative to their peers who do not become pregnant (Furstenberg, Brooks-Gunn, and Chase-Lansdale, 1989), perhaps explaining the positive associations between premarital pregnancy and subsequent risk of divorce (Furstenberg, 1976). But research in this vein does not account for the relational context in which those pregnancies occurred.
To the extent that adolescents’ experiences in romantic relationships have the immediate consequences reviewed in the previous section, it is reasonable to expect that evidence of long-term consequences may be found, because those immediate consequences have often been linked to adult marital outcomes. For example, depression in adolescence, which is predicted by the breakup of adolescent romantic relationships (Monroe et al., 1999), predicts earlier entry into marriage and higher risk of divorce in early adulthood (Gotlib, Lewinsohn, and Seeley, 1998). Intimate partner violence among young adults, which is predicted by violent behavior in adolescent relationships (Capaldi et al., 2001), is one of the strongest predictors of divorce in the early years of marriage (Rogge and Bradbury, 1999). The fact that these sorts of variables have been associated with adult marital outcomes in correlational and longitudinal research is consistent with models that view adolescent relationships as central to the foundation of healthy adult marriages.
Yet without direct evidence that rules out alternative explanations of the obtained associations, conclusions about the role of adolescent relationships in adult marriages must remain tentative. Research on the long-term implications of peer relationships provides a cautionary tale. In 1982, Giordano and colleagues conducted interviews with 942 adolescents regarding their relationships with their friends. Ten years later, the researchers recontacted 620 of their original respondents to examine whether aspects of peer relationships accounted for variance in a range of adult outcomes, controlling for economic and demographic variables and relationships with parents (Giordano et al., 1998). Analyses revealed that the quality of adolescent’s relationships with peers did not account for significant variability in adult outcomes after sociodemographic and parental variables were controlled. However, sociodemographic and parental variables did account for adult outcomes, including marital satisfaction, after controlling for the quality of adolescent peer relationships.
Similar analyses of the long-term implications of adolescent romantic relationships are crucial to determine (1) whether associations between adolescent romantic experiences and adult marital outcomes exist, and (2) whether those associations are more than simply the byproduct of the common effects of sociodemographic variables and the family of origin.
Summary and Conclusions
As Collins and Madsen (2006) recently observed,
The agenda for filling gaps in research on relationships during early adulthood is a lengthy one. . . . the most compelling accounts would come from longitudinal data sets in which repeated accounts are sought from the same individuals across three age periods, using standard reporting devices and using standard metrics (p. 24).
Research fitting this description is being conducted, but the results of this work are not yet available. This leaves the available data to inform efforts to intervene in adolescent romantic relationships incomplete in several important ways.
First, as noted at the outset, the methods employed in much of the existing research have been quite limited. This report has emphasized the best of the available research, studies that have incorporated multiple assessments of individuals over time, multiple reporters (including the individual, parents, and peers), and a broad range of variables. However, the majority of this research has collected data at only a single occasion, relied exclusively on individuals’ self-reports, and assessed only a few variables at a time. Importantly, even the strongest studies in this area have sampled almost exclusively from populations that are predominately middle-class and white and have not followed these samples beyond the earliest years of young adulthood.
Recognizing the limitations of the sampling in most research on adolescents has led some to question whether the results of the existing literature can be applied to nonwhite or low-income populations (McLoyd and Steinberg, 1998). Until more research is conducted on ethnically diverse and low-income samples, or until results from nationally representative studies are broken down by income, this remains an open question. Recognizing the limitations of the study interval has led others to question whether the existing research demonstrates a clear link between adolescent relationships and adult marital outcomes. The absence of data directly linking adolescents’ experiences in romantic relationships to the outcomes of their marriages in adulthood leaves ongoing efforts to intervene in adolescent romantic relationships resting on an uncertain empirical foundation.
Second, despite the range of variables that have been addressed by this literature, broad classes of variables have nevertheless been overlooked. For example, whereas there have been several studies examining the timing of adolescents’ entry into romantic relationships, the quality of those relationships, and processes within those relationships, have received far less attention (Collins and Van Dulmen, 2006). As a result, the ways that adolescents maintain their relationships, and the determinants of relationship satisfaction at different ages, remain unknown. Although most adolescents have multiple romantic relationships before they become young adults, the development of an individual across relationships, and the degree of consistency between relationships, has also not been studied (Bouchey and Furman, 2003). Research in this area has focused on the implications of being in relationships, rather than the implications that a specific relationship may have for an individual. Finally, given the emphasis on beliefs and values about marriage in current interventions targeting adolescent romantic relationships, it is noteworthy how seldom researchers studying these relationships have directly examined the implications of adolescents’ beliefs and values about marriage. These variables have been described as dependent variables but not as independent variables that may account for subsequent adolescent outcomes. Each of these gaps in the existing literature represents a potentially fruitful direction for future research in this area.
Third, researchers in this area have tested specific hypotheses infrequently, and they have pitted alternative hypotheses or models against each other even less frequently. As a result, research has not accumulated to inform, refine, or elaborate upon theoretical frameworks, and competing frameworks have not emerged to guide research. The significant consequence of this limitation is a general failure to examine threats to causal inferences throughout the literature. Too often, cross-sectional or longitudinal associations among variables of interest have been interpreted as support for models that consider only those variables (e.g., associations between relationships with parents and with romantic partners during adolescence have been taken as support for attachment perspectives).Yet these same analyses often overlook variables (such as socioeconomic status) that do not feature in those models but that might nevertheless influence both variables. Without such analyses, well-supported causal statements remain elusive.
Yet despite these significant limitations, the existing research on the antecedents and consequences of adolescent romantic relationships offers some justification for current efforts to target these relationships for intervention. Although no single study has yet addressed the entire framework, the results of correlational and longitudinal studies are generally consistent with specific paths of the integrative framework described in Chapter Three. Characteristics of children’s family of origin and early environment are associated with their peer relationships. Peers do appear to influence timing of entry into romantic relationships. The immediate consequences of these relationships are the very ones that would be expected to influence adult marital outcomes. Moreover, studies that have followed individuals from early childhood through young adulthood do point to adolescence as a critical period during which significant antecedents of young adult outcomes take shape (e.g., Zimmer-Gembeck, Siebenbruner, and Collins, 2004). Thus, the existing data are consistent with a model that views adolescent romantic relationships as a key period during which the foundations of healthy adult marriages may be strengthened. To the extent that further research confirms this emerging picture of adolescent relationships as a link in a causal chain leading from the early environment to adult marriages, it would be warranted to intervene during adolescence, especially as a means of preventing the continuation of negative patterns in vulnerable youth. How are currently available interventions attempting to accomplish this? We turn to that question in the next chapter.
| Table of Contents | Previous | Next |