Romantic Relationship Churn in Early Adolescence Predicts Hostility, Abuse, and Avoidance in Relationships Into Early Adulthood

Abstract

This study examined early adolescent romantic “churning,” defined here as having a large number of boyfriends/girlfriends by age 13, as a problematic marker likely to predict hostility, abuse, and avoidance during conflict in later relationships. A sample of 184 adolescents was followed through age 24 to assess predictions of hostility, abuse, and avoidance during conflict from early romantic churning. Controlling for gender and family income, romantic churning at age 13 predicted relative decreases in peer preference and relative increases in conflict and betrayal in close friendships from ages 13–16, as well as higher observable hostility and self and partner-reported abuse in romantic relationships by age 18 and greater avoidance during conflict with romantic partners by age 24. Findings remained after accounting for attachment security, social competence, and friendship quality in early adolescence, suggesting that early romantic churning may uniquely predict a problematic developmental pathway.

Do early adolescent romantic relationships really matter? As important as they often seem to teens, these relationships are frequently considered by adults to be both relatively inconsequential in nature at the time and of minimal long-term significance (Meier & Allen, 2009; Roisman, Masten, Coatsworth, & Tellegen, 2004). For many adolescents, this stereotype may be completely accurate; yet, prior research and theory suggests that the potential exists for these relationships to lead down far more problematic developmental paths (Brown, 1999; Collins, 2003; Furman & Shaffer, 2003). Early adolescence is a time of establishing identities and learning how to navigate increasingly central peer relationships (Brown, 1999; Furman & Shaffer, 2003). According to Brown (1999), early adolescence tends to be characterized by the “initiation” phase of romantic relationships, when few adolescents report dating more than rarely, they tend to have few partners, and they are focused on developing an identity as someone with romantic interests. In contrast to this typical developmental path, overinvolvement in romantic experiences in early adolescence may be associated with long-lasting problems.

More specifically, the biologically and socially driven intensity of romantic relationships has the potential to lead some young teens to continually, and even obsessively, seek out such relationships while, given their relative immaturity, finding they are over and over again unable to sustain relationships for significant lengths of time. This over-involvement and experience of repeated churn (here defined as experiencing a large number of romantic relationships in a short period of time—an early adolescent analogue of the on-again off-again relationship pattern that has been identified later in development (Halpern-Meekin, Manning, Giordano, & Longmore, 2013)) has the potential to alter development in three key ways. It is likely to distract teens from the more developmentally stage-salient task of forming close same-gender friendships; to lead teens to devalue relationships that are constantly changing; and to lead to significant distress for teens. Unfortunately, each of these patterns is ultimately likely to lead to problematic long-term relationship patterns, although this possibility has received scant attention in the study of adolescent romantic relationship development to date.

First, the intensity of early romantic relationships, in terms of the adolescents’ internal experience, if not societal judgments, renders them likely to distract from what we are increasingly learning is a core task of this age—establishing strong same-gender friendships (Fraley & Roisman, 2015). Romantic relationships in early adolescence are typically embedded in the broader peer group and are not primary at this developmental stage (Brown, 1999). Churning through romantic interests undoubtedly requires both time and emotional energy in this pursuit, which in turn is likely to crowd out other pursuits (Laursen & Williams, 1997; Zimmer-Gembeck, 1999). We then see a pattern that is the opposite of Harry Stack Sullivan’s now well-supported notion of the developmental importance of ‘chumship’ at this age (Berndt, 1982; Sullivan, 1953; Waldrip, Malcolm, & Jensen‐Campbell, 2008).

Romantically overinvolved adolescents may become less desirable as non-romantic companions and develop more problematic peer relationships over time as they continually prioritize romantic interests. Little research has examined this, although we know that, for already unpopular early adolescents, romantic involvement is associated with poorer emotional and behavioral adjustment (Brendgen, Vitaro, Doyle, Markiewicz, & Bukowski, 2002). The social lives of adolescents who churn through romantic interests may also inherently be perceived as “high drama,” with frequent relationship formations and dissolutions. Although such drama may be fascinating to early adolescents, over time it seems likely to become less attractive within the broader peer group as these adolescents likely prioritize romantic interests, neglect friendships, and fail to develop relationship skills in the context of these friendships. For example, we know from previous research that the link between “pseudomature” behavior (including precocious low level romantic involvement) and popularity fades over time (Allen, Schad, Oudekerk, & Chango, 2014).

In addition, overinvolved adolescents may come to devalue individual relationships as they experience constant turnover within romantic interests. As these adolescents develop expectations that relationships are inherently transitory and begin to act on these expectations, they may become less desirable potential friends and partners (Downey, Freitas, Michaelis, & Khouri, 1998) and as a result, develop more hostile and conflictual friendships and romantic relationships over time (Loeb, Tan, Hessel, & Allen, 2018). Adolescents who expect relationships to end may behave defensively or hostilely in response to those expectations, which may reciprocally elicit hostility from partners (Downey et al., 1998). We also know that friendship competence can act as a buffer against some of the negative repercussions of romantic involvement in adolescence (Szwedo, Chango, & Allen, 2015). Yet, adolescents who continually focus on romantic relationships are likely to experience more of the unbuffered negative effects of romantic relationship stressors and may develop less competent romantic relationships. This is consistent with work that has found that early adolescents classified as “Early Starters” in terms of romantic involvement developed more externalizing problems and that romantic overinvolvement in late adolescence is associated with delinquency and relative decreases in emotional health (Connolly, Nguyen, Pepler, Craig, & Jiang, 2013; Cui, Ueno, Fincham, Donnellan, & Wickrama, 2012; Zimmer-Gembeck, Siebenbruner, & Collins, 2001), although these studies did not examine the long-term implications of early romantic churning.

Next, almost by definition, experiencing a large number of romantic relationships also means experiencing a large number of relationship dissolutions. For older adolescents and young adults, relationship dissolution is associated with multiple signs of distress, including sadness, the onset of major depression, anger, substance use, and a reduced sense of self-concept (Laursen, Finkelstein, & Betts, 2001; Monroe, Rohde, Seeley, & Lewinsohn, 1999; Sbarra, 2006). Although relationship dissolution at 13 is likely to have different meanings than it does at older ages, the constant churn in the romantic lives of early adolescents with a history of multiple romantic relationships tells us something important: These adolescents clearly both desire romantic involvement and are unable to sustain it for any length of time. This repeated romantic involvement is likely to overwhelm adolescents’ fledgling capacities to regulate emotion (Davila et al., 2009; Gullone, Hughes, King, & Tonge, 2010; Joyner & Udry, 2000; Larson, Clore, & Wood, 1999). This, in turn, is likely to lead to increased anger and hostility in relationships as adolescents struggle to effectively manage intense negative emotions and expectations of rejection or loss that are likely to arise within these relationships (Romero‐Canyas, Downey, Berenson, Ayduk, & Kang, 2010).

Although adolescents who churn through romantic interests in early adolescence may develop hostility and conflict within relationships, other outcomes may potentially be even more problematic. In line with the theory of developmental cascades (Masten & Cicchetti, 2010), as development progresses, adolescents who churn through romantic interests may find themselves in a cascading series of different types of relationships in which conflict and negative affect are handled poorly. This is similar to the finding that deviancy training in adolescent friend relationships predicts coercion in adult romantic relationships (Ha, Otten, McGill, & Dishion, 2019). Romantic churning likely increases the risk of developing coercive or even abusive relationships as adolescents age, as they forgo the opportunity to develop more adaptive conflict resolution skills normally honed in friendships, and do not maintain a romantic relationship long enough to learn to navigate conflict within that relationship (Laursen et al., 2001; Mulford & Giordano, 2008). Some evidence supporting this idea comes from short-term, self-report research finding higher levels of victimization for adolescents (ages 12–21) who had more than one romantic partner in the previous 18 months (Halpern, Oslak, Young, Martin, & Kupper, 2001). Although this short-term study only peripherally addressed the issue of significant relationship churn and did not focus on early adolescence, it nonetheless suggests the plausibility of the idea that relationship turnover is linked to conflict and abuse in relationships. Although romantic involvements in early adolescence can appear transitory and inconsequential, finding links to later abusive behavior would suggest that romantic churning in early adolescence may serve as an early warning sign of a risk for more serious later problems.

Hostility and abuse in relationships tend to peak in the early twenties and then decrease with age (Rennison & Welchans, 2000). By early adulthood, early adolescents who have churned through romantic interests may find themselves in relationships that are less openly hostile, but are instead characterized by greater avoidance and less positivity, particularly during conflict. We know that intimate partner violence is associated with a demand-withdraw pattern (Babcock, Waltz, Jacobson, & Gottman, 1993; Holtzworth-Munroe, Smutzler, & Stuart, 1998) and previous research has found that conflict avoidance and withdrawal are associated with relationship dissatisfaction and less intimacy (Heavey, Christensen, & Malamuth, 1995; Knobloch & Carpenter-Theune, 2004; Smith, Heaven, & Ciarrochi, 2008). If adolescents continually face problematic and hostile relationships, they may ultimately develop an avoidant style of interacting and experience emotionally distant romantic relationships in adulthood. Yet no research to date has examined avoidance as a possible long-term consequence of early romantic churning.

Although there are a number of paths by which romantic churn in early adolescence could lead to future dysfunction, it may also serve as an important early marker of broader patterns of intrapersonal dysfunction. In other words, adolescents who churn through romantic interests may be a select group with a number of risk factors, reflecting ongoing problems that result in churn and other negative outcomes. One way to distinguish between these alternative explanations of links between churning and later dysfunction is to examine the extent to which churning is predictive over and above other early markers of relationship competence. Three markers that seem likely to be implicated in both romantic churning and later relationship outcomes are attachment insecurity, lack of social competence, and poor close friendship quality. Attachment security has been shown to be linked to relationship qualities across development (Hazan & Shaver, 1987; Simpson, 1990; Van IJzendoorn, 1995) and seems likely to be involved in early adolescents’ approach to romantic relationships as well as friendships. In addition, adolescents who are less socially competent may struggle to maintain developmentally appropriate relationships and may also experience more problematic relationships over time (Buhrmester, Furman, Wittenberg, & Reis, 1988; Conger, Cui, Bryant, & Elder Jr, 2000). Finally, those who begin adolescence with more problematic close friendships may continue to struggle to manage close relationships across development (Bagwell, Newcomb, & Bukowski, 1998; Bukowski, Newcomb, & Hartup, 1998). The current study examined each of these markers as potential explanatory factors for any observed effects of romantic churning. In addition, we explored possible moderation by family income and gender, as previous research suggests that romantic experiences vary according to these factors (Ciairano, Bonino, Kliewer, Miceli, & Jackson, 2006). Wherever possible in analyses, we also used reports or observations from both participants and friends/partners to better understand how early romantic churning may be associated with later relationship problems.

This study used longitudinal, multi-method data in a community sample of adolescents followed from age 13 to age 24 to examine links between early adolescent overinvolvement and churning through romantic relationships and later problems in peer and romantic relationship functioning. It was specifically hypothesized that:

1. The number of boyfriends/girlfriends reported would be associated with other markers of early romantic overinvolvement.

2. High levels of romantic churn at 13 would predict decreasing preference by peers and increasingly conflictual friendships from age 13 to age 16.

3. Romantic churning at age 13 would predict more hostile behaviors during conflict and abuse within romantic relationships at age 18 and more avoidance during conflict in romantic relationships age 24.

4. Conflictual friendships from ages 14–16 would mediate the relationships between romantic churning at 13 and later romantic relationship outcomes.

5. Romantic churning would predict problematic relationship outcomes over and above baseline indicators of social functioning.

Methods

Interview and observational data were obtained longitudinally for a community sample of 184 adolescents (86 males, 98 females) first assessed at age 13 and re-interviewed annually up through age 24. Close friends and late adolescent and early adult romantic partners also participated. The sample was demographically diverse (baseline median family income was in the $40,000 to $59,000 range; the sample was 63% European-American, 27% African-American, and 14% mixed race or other race/ethnicities). Adolescents were originally recruited from the seventh and eighth grades at a public middle school drawing from suburban and urban populations in the Southeastern United States. Students were recruited via an initial mailing to all parents of students in the school, along with follow-up contact efforts at school lunches. Adolescents who indicated they were interested in the study were contacted by telephone. Of all students eligible for participation, 63% agreed to participate either as target participants or as peers providing collateral information.

For the purposes of the present study, data were drawn from six time points: First, in early adolescence with participants and their close friends (participant M age = 13.35, SD = 0.64); next, three times in mid-adolescence with participants and their close friends (participant M age = 14.27, SD = .77; participant M age = 15.21 SD = .81; and participant M age = 16.35, SD = .87); next, at a late adolescent assessment of target participants and their romantic partners (participant M age = 18.33, SD = 0.95), and finally at an early adulthood assessment of the target participants and romantic partners (participant M age = 23.78, SD= 0.97).

At ages 13 and 14–16, participants nominated a close friend to participate in the study. Close friends were defined as “people you know well, spend time with, and whom you talk to about things that happen in your life.” These close friends were chosen by the participant and may not have reciprocally considered the target participant to be their closest friend. At age 13, close friends reported that they had known the participants for an average of 4.02 years (SD = 2.88). At age 14, close friends reported knowing the participants for an average of 4.27 years (SD = 3.09). At age 15, close friends reported knowing the participants for an average of 5.07 years (SD = 3.41). At age 16, close friends reported knowing the participants for an average of 5.72 years (SD = 3.82). Twenty participants (10.87%) brought the same friend to all four assessment points.

At the age 18 and 24 assessments, participants in a romantic relationship of at least 3 months were invited to participate in filmed interaction tasks with their romantic partners. To maximize the number of romantic partners able to participate, dyads came in once over a span of three years to complete observational and questionnaire measures. At the age 18 data collection, seventy-five (40.76%) of the original participants were in eligible romantic relationships and both they and their partners agreed to participate. Participants reported knowing their romantic partners an average of 1.22 years (SD = 1.13 years). At the age 24 data collection, 88 dyads (47.83% of the original sample) were eligible and agreed to participate. Participants reported knowing their romantic partners an average of 4.66 years (SD = 3.77 years). Two participants (2.86%) brought the same romantic partner to the age 18 and age 24 data collection.

Of those participants who did not participate at the age 18 data collection, 75 (68.81%) did not meet the criteria of being in a relationship lasting at least three months. Of the particpants who did not participate in the age 24 data collection, 57 (59.38%) did not meet the criteria of being in a relationship lasting at least three months. For the remainder in both cases, the majority of cases of non-participation were a result of partners’ declining our invitation to participate and/or the inability to schedule an observational assessment in which both parties were willing and able to participate. Analyses indicated that there were no significant differences between those who did vs. did not participate at ages 18 or 24 on gender, family income, or initial levels of the variables measured.

To best address any potential biases due to attrition and missing data in longitudinal analyses, full information maximum likelihood methods were used, with analyses including all variables that were linked to future missing data (i.e., where data were not missing completely at random). Because these procedures have been found to yield less biased estimates than approaches (e.g., simple regression) that use listwise deletion of cases with missing data, the entire original sample of 184 for the larger study was utilized for these analyses. This analytic technique does not impute or create any new data nor does it artificially inflate significance levels. Rather, it simply takes into account distributional characteristics of data in the full sample so as to provide the least biased estimates of parameters obtained when some data are missing (Arbuckle, 1996).

Measures

Number of Boyfriends/Girlfriends (Age 13).

At age 13, participants responded to the question “How many boyfriends or girlfriends have you had in total?” Responses ranged from 0–36, with a mean of 5.54, standard deviation of 6.50, a median of 3.00 and a mode of 0.00. More than half (53.22%) of participants reported having had 3 or fewer boyfriends or girlfriends. To account for the possibility that outliers were driving observed associations, outliers three or more standard deviations from the mean (i.e. 25 boyfriends or girlfriends or greater) were changed to within three standard deviations of the mean (Winsorization) using the procedure described in (Reifman & Keyton, 2010). After this adjustment, the mean number of boyfriends/girlfriends was 5.43 and the standard deviation was 6.11. Results did not substantively change using the unadjusted vs. adjusted variable.

Early Romantic Involvement (Age 13).

At age 13, participants responded to a series of questions about romantic involvement. We first asked: “Have you ever gone out with a boy or girl? (This includes going somewhere with them; asking or being asked to have a special relationship with a boy or girl).” Participants selected yes or no. If they selected yes, they were instructed to choose from a list of activities they might have done with this person. Next they responded to the question “Did you do these things more as just the two of you (A) or in a group of friends (B)?” They then rated how many people they had held hands with (either while walking or while sitting), how many people they had kissed on the lips, and how many people they had “made out with (kissed a long time and touched),” all on a scale from “none” to “more than 10.” Finally, they responded to the question: “Have you ever had sex (‘gone all the way’)?”

Peer Preference (Ages 13 and 16).

Peer preference was assessed at ages 13 and 16 using a limited nomination sociometric procedure. Each adolescent, his or her closest friend, and two target peers named by the adolescent were asked to nominate up to 10 peers in their grade with whom they would “most like to spend time on a Saturday night.” This measure is similar to one created by Franzoi, Davis, & Vasquez-Suson, 1994. The raw number of “like” nominations each teen received was standardized within grade level before being added to the main data set as the primary measure of preference following the procedure described in Coie, Dodge, & Coppotelli, 1982. Such ratings are not the only relevant marker of competence with peers (e.g., perceived status with peers rather than likeability, or combinations of liked and disliked nominations; Coie, Dodge, & Kupersmidt, 1990; Prinstein, 2007), however, simple sociometric ratings of likeability do capture one of the most direct markers of ability to relate positively to a broader peer group (Prinstein, 2007). This procedure resulted in a sample of 72 – 146 teens (depending on the grade level), comprising approximately 38% of the entire student population in these grades, who provided nominations of anyone in their grade at school. Though this method may somewhat inflate the number of “like” nominations each participant received (by using a very small percentage [i.e., 2 – 5%] of nominators selected by the participant in the sample of raters), this effect would apply to the same degree for all participants and thus is not likely to substantively affect results. Grade-based nominations were utilized rather than classroom-based nominations due to the age and classroom structure of the school that all of the participants attended. The large number of raters for each teen means that this subsample of nominators is likely to yield fairly reliable estimates of preference for each teen (Prinstein, 2007). This procedure has been validated in past studies with this same sample, with social preference being linked to numerous positive indicators of adjustment, including attachment security, ego development, and competence in close friendships (Allen, Porter, McFarland, McElhaney, & Marsh, 2007; Allen, Porter, McFarland, Marsh, & McElhaney, 2005) as well as predictive of functioning into adulthood (Narr, Allen, Tan, & Loeb, 2017).

Attachment States of Mind (Age 14).

Participants completed the Adult Attachment Interview (AAI; George, Kaplan, & Main, 1996), a structured interview designed to capture states of mind regarding attachment. The interview took about one hour and probed participants’ descriptions of their relationships with their parents. Some adaptations to the adult version were made to make the questions more appropriate for an adolescent population (Ward & Carlson, 1995). Interviews were audiotaped and transcribed for coding. Interviews were then coded using the AAI Q-set (Kobak, Cole, Ferenz‐Gillies, Fleming, & Gamble, 1993) by at least two raters trained in the Q-sort and the Adult Attachment Interview Classification System. Each rater read a transcript and used a forced distribution to provide a Q-sort description by assigning 100 items into nine categories ranging from most to least characteristic of the interview. The Q-sorts were then compared to a prototype sort for a maximally secure transcript, which was then used as that participant’s scale security score (ranging from −1.00 to 1.00). The Spearman-Brown interrater reliability for the final security scale score was .82.

Friendship Quality (Age 13).

The Friendship Quality Questionnaire (Parker & Asher, 1993) is a 40-item self-report measure designed to capture adolescents’ and friends’ reports of friendship adjustment with their identified closest friend. Participants and their closest friends responded on a 5-point Likert scale from “Not at all true” to “Really true.” Their scores were then averaged to capture dyadic perceived friendship quality. Example items include “We always spend free time at school with each other” (Companionship and recreation) and “We talk about how to get over being mad at each other” (Conflict resolution). Internal consistency was in the “excellent” range (Cronbach’s α=.95).

Conflict with Friends (Age 13 and 14–16).

The Friendship Quality Questionnaire (Parker & Asher, 1993) is a 40-item self-report measure designed to capture adolescents’ and friends’ reports of friendship adjustment. Participants and friends responded on a 5-point Likert scale from “Not at all true” to “Really true.” The current study uses the 7-item Conflict and Betrayal subscale (which is reverse-coded so that higher scores mean less conflict and betrayal). Example items include “We get mad at each other a lot” and “We can count on each other to keep promises” (reverse scored). To create the 14–16 composite measure of conflict and betrayal, participants’ and friends’ scores were averaged together across the time period, in order to capture dyadic perceptions of conflict. The average internal consistency was in the “acceptable” range (Cronbach’s α=.70).

Social Competence (Age 13).

The Adolescent Problem Inventory has been used as a measure of adolescents’ social problem-solving competence and to probe teens’ expectations and values about competent behaviors (Freedman, Rosenthal, Donahoe Jr, Schlundt, & McFall, 1978; Gaffney & McFall, 1981). Scenarios were read to teens and they were followed by an incompetent response or a competent response. An example item includes: “Regina has been your best friend for years but lately you’ve been hanging out sometimes with a new friend named Shari. Your aunt is going to take you to [a theme park] and will let you invite one friend. You invite Shari. Regina finds out and is really upset. She comes up to you at school and asks why you didn’t invite her.” They were then given the competent prompt: “When some teens heard that story, they told us: I’d tell Regina that I still want to be friends but make sure she knows I have other friends too.” The teen’s responses were coded to see how closely they matched the hypothetical teen response in order to see how influenced they were by hearing the opinion of a competent teen. For the current study, the total amount that participants were competently influenced was used as a measure of social competence (i.e., how similar their responses were to a socially competent response). This measure has been used to capture social competence and social problem solving in several previous studies (Allen, Chango, Szwedo, Schad, & Marston, 2012; Allen et al., 2002; Kuperminc & Allen, 2001). Interrater reliability was calculated for the competent influence scale using intraclass correlation coefficients and was in what is considered “excellent” range for this statistic (intraclass r = .93; Cicchetti & Sparrow, 1981).

Observed Hostile Behaviors in Relationships (Ages 13 and 18).

At age 13, participants and their close friends participated in a revealed differences task in which they had to come to a consensus on a hypothetical task (which people should be given a place on a rescue shuttle from Mars). At age 18, participants and their romantic partners participated in a similar revealed differences task. All interactions lasted 8-minutes and were video recorded for coding purposes. The coding system employed yields ratings for each participant’s overall behavior toward his/her partner in the interaction (Allen, Hauser, Bell, & McElhaney, 2000; Allen, Hauser, Eickholt, Bell, & O’Connor, 1994). The scale includes: 1) Overpersonalizing behaviors: Treating the disagreement as being in some respect a “fault” or feature of the person disagreeing rather than a difference in ideas and reasons; 2) Pressuring behaviors: The extent to which the individual proceeds in the discussion as though his/her main objective is to get his/her own selections accepted; 3) Avoidance behaviors: The degree to which an individual steers away from disagreements or the chance to clarify disagreements; and 4) Rudeness: The use of hostile comments, interruptions, or other tactics that undermine the relationship. Higher scores represent higher observed hostility. Participants’ and partners’ scores were averaged together to capture dyadic level hostile behaviors. Interrater reliability was calculated for the overall scale using intraclass correlation coefficients and was in what is considered “good” range for this statistic (intraclass r = .69-.83; Cicchetti & Sparrow, 1981).

Abuse/Coercion with Romantic Partners (Age 18).

The Conflict in Relationships scale (Wolfe, Reitzel-Jaffe, Gough, & Wekerle, 1994) was created to measure physical and verbal abuse and positive and negative communication patterns in dating situations that have been experienced or committed by the participant. The current study uses the 15-item Abuse/Coercion subscale. Both participants and partners reported on the amount of abuse/coercion they perpetrated towards their partner during a conflict/argument in the past year and their scores were averaged together to capture dyadic level abuse/coercion. Example items include: “I tried to turn my partner’s friends against him/her,” and “I threatened my partner in an attempt to have sex with him/her.” Internal consistency was in the “acceptable” range (Cronbach’s α=.73 to .77). Because this variable was positively skewed, a log-transformation was used.

Observed Avoidance in Conflicts (Ages 13 and 24).

Participants and friends (age 13) or romantic partners (age 24) completed the same 8-minute, video recorded conflict task described above. The scale included the following behaviors: stating reasons for a position, confidence, positive connectedness and collaborative behavior. Interrater reliability was calculated for the overall scale using intraclass correlation coefficients and was in what is considered the “good” to “excellent” range for this statistic (Intraclass r = .66-.91; Cicchetti & Sparrow, 1981). Dyadic-level avoidance during conflict was created by reverse-scoring the measure of positive behaviors, to capture the extent to which participants’ interactions were characterized by a lack of these assertive behaviors.

Results

Preliminary Analyses & Descriptive Statistics

Table 1 presents means, standard deviations, and intercorrelations of substantive variables. Gender and family income were correlated with several variables of interest; hence, these demographic factors were assessed as covariates in all analyses below. We also examined the possible moderating effects of gender and family income on each of the relations described in the primary analyses below. All moderating effects analyzed were obtained by creating interaction terms based on the product of centered main effect variables. No moderating effects were found.

Table 1.

Correlations Among and Descriptive Statistics For Key Study Variables

	M (SD)	N	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.
1. BF/GF (13)	5.43 (6.11)	171	--	−03	−24**	20**	38***	02	27*	38***	21**	30**	−27***	−05
2. Peer pref. (13)	0.96 (1.35)	184		--	56***	03	08	−05	−20	−07	−33***	−43***	32***	04
3. Peer pref. (16)	.42 (1.09)	151			--	04	23**	−03	−08	−24*	−28***	−37**	39***	−09
4. Conf./betray (13)	31.33 (3.00)	184				--	35***	−16*	−06	−09	−02	−27**	18**	04
5. Conf./betray (14–16)	31.33 (2.40)	181					--	−21**	−13	−48***	−07	11	12	32***
6. Obs. hostility (13)	.77 (.48)	178						--	45***	17	01	−09	−02	−05
7. Obs. hostility (18)	.76 (.44)	75							--	28*	32**	23	−22	02
8. Abuse/coercion (18)	2.89 (.15)	92								--	24*	−04	−22*	−03
9. Avoidance (13)	1.63 (.60)	178									--	21*	−31***	08
10. Avoidance (24)	1.64 (.45)	88											−31**	19
11. Family income	43,600 (22,4000)	181											--	−11
12. Gender	--	184												--

^*p < .05.

^**p < .01.

^***p <.001.

Primary Analyses

Hypothesis 1: The number of boyfriends/girlfriends reported will be associated with other markers of early romantic overinvolvement.

To test this hypothesis, we examined romantic activity at age 13 in the current sample. See Table 2.

Table 2.

Descriptive Information about Romantic Behavior and Correlations with Number of Boyfriends/Girlfriends

	Yes (1)	No (0)				Correlation with number of boyfriends/girlfriends
Have you ever gone out with a boy or girl?	137 (74.9%)	46 (25.1%)				--
Have you ever gone with someone you considered a boyfriend or girlfriend?	127 (92.7%)	10 (7.3%)				--
Have you ever had sex?	15 (8.4%)	164 (91.6%)				.41***
	Just the two of you (1)	In a group of friends (0)
Do you spend time…	56 (41.5%)	79 (58.5%)				.24**
	0	1–2	3–5	6–10	More than 10
How many people have you held hands with?	40 (22%)	41 (22.5%)	43 (23.6%)	24 (13.2%)	34 (18.7%)	.48***
How many times have you kissed on the lips?	64 (35.6%)	37 (20.6%)	31 (17.2%)	17 (9.4%)	31 (17.2%)	.48***
How many people have you “made out with?”	106 (59.6%)	37 (20.8%)	20 (11.2%)	10 (5.6%)	5 (2.8%)	.35***

^*p < .05.

^**p < .01.

^***p <.001.

The number of boyfriends/girlfriends reported was correlated with the number of people with whom adolescents had held hands (r=.48, p=.001), as well as the number of people kissed on the lips (r=.48, p=.001) and the number of people with whom adolescents had made out (r=.35, p=.001). Adolescents who reported having more boyfriends or girlfriends were significantly more likely to have had sex (t=4.98, p=.001). Those who had not had sex reported an average number of 4.68 boyfriends or girlfriends (SD=5.83), while those who had had sex reported an average number of 13.57 boyfriends or girlfriends (SD=6.57).

Those who said they mainly spent time “in a group of friends” had, on average, 6.22 boyfriends or girlfriends (SD=5.43), while those who mainly spent time “just the two of [them]” had an average of 9.30 boyfriends or girlfriends (SD=7.47). This was a significant difference (t=−2.55, p=.01). These findings suggest that having a greater number of boyfriends/girlfriends by age 13 is, in fact, associated with other markers of precocious romantic involvement in adolescence.

Next, regressions using Full Information Maximum Likelihood analyses were conducted in MPlus (Version 7.2;(Muthén & Muthén, 2015).

Hypothesis 2: High levels of romantic churn at 13 will predict decreasing preference by peers and increasingly conflictual friendships from age 13 to age 16.

As shown in Table 3, controlling for participant gender, family income, and peer preference at age 13, a greater number of boyfriends/girlfriends by age 13 predicted relative decreases (i.e., accounting for baseline peer preference) in peer preference by age 16 (β =−.21, p=.003). This suggests that, apart from demographic variables, having a greater number of boyfriends/girlfriends at age 13 was predictive of decreasing preference by the broader peer group by age 16.

Table 3.

Predicting Peer Preference (Age 16)

	β	Δ R2	Total R2
Step 1.
Gender (1=M; 2=F)	−.09
Family Income	.16*
			.155**
Step 2.
Peer Preference (13)	.51***
		.216***	.371***
Step 3.
Number of Boyfriends/Girlfriends (13)	−.21**
		.039**	.410***

Note.

^*p <.05.

^**p ≤.01.

^***p ≤.001. β’s are from final model.

As shown in Table 4, controlling for gender, income, and conflict/betrayal at age 13, target participants’ reports of having more boyfriends/girlfriends by age 13 predicted relative increases in conflict/betrayal from ages 14–16 as reported by target participants and close friends (β =.30, p=.001). This suggests that having more boyfriends/girlfriends by 13 was predictive of increasingly conflictual friendships, even accounting for demographic variables and baseline levels of conflict and betrayal.

Table 4.

Predicting Conflict and Betrayal in Friendships (Ages 14–16)

	β	Δ R2	Total R2
Step 1.
Gender (1=M; 2=F)	.31***
Family Income	.03
			.130**
Step 2.
Conflict and Betrayal (13)	.27***
		.082***	.212***
Step 3.
Number of Boyfriends/Girlfriends (13)	.30***
		.061***	.273***

Note.

^*p <.05.

^**p ≤.01.

^***p ≤.001. β’s are from final model.

Hypothesis 3: Romantic churning at age 13 will predict more hostile behaviors during conflict and abuse within romantic relationships at age 18 and more avoidance during conflict in romantic relationships age 24.

As shown in Table 5, controlling for gender, income, and observed hostility with a close friend at age 13, target participants’ reports of having more boyfriends/girlfriends by age 13 predicted more observed hostile behaviors from both partners at age 18 (β =.29, p=.03). These results suggest that having more boyfriends/girlfriends by age 13 was predictive of observable hostility within a romantic relationship at age 18, even taking into account friendship hostility at age 13.

Table 5.

Predicting Observed Hostility in Romantic Relationships (Age 18)

	β	Δ R2	Total R2
Step 1.
Gender (1=M; 2=F)	.02
Family Income	−.10
			.038
Step 2.
Observed Hostility (13)	.43***
		.193***	.231**
Step 3.
Number of Boyfriends/Girlfriends (13)	.29*
		.066*	.297***

Note.

^*p <.05.

^**p ≤.01.

^***p ≤.001. βs are from final model.

As shown in Table 6, controlling for gender, income, and observed hostility with a close friend at 13, target participants’ reports of having more boyfriends/girlfriends by age 13 predicted more reported abuse/coercion in the relationship at 18 (β =.32, p=.001). These findings suggest that having more boyfriends/girlfriends by 13 predicted higher levels of abuse from both participants and partners by 18, even accounting for demographic variables and baseline levels of observed hostility within a close friendship.

Table 6.

Predicting Abuse/Coercion in Romantic Relationships (Age 18)

	β	Δ R2	Total R2
Step 1.
Gender (1=M; 2=F)	−.01
Family Income	−.11
			.041
Step 2.
Observed Hostility (13)	.13
		.024	.065
Step 3.
Number of Boyfriends/Girlfriends (13)	.32***
		.091**	.156*

Note.

^*p <.05.

^**p ≤.01.

^***p ≤.001. βs are from final model.

As shown in Table 7, controlling for gender, income, and observed avoidance within a close friendship at 13, target participants’ reports of having more boyfriends/girlfriends by age 13 predicted more observed avoidance within a romantic relationship at age 24 (β =.30, p=.002). These findings suggest that having more boyfriends/girlfriends by 13 predicted avoidance during conflict by age 24 within a romantic relationship, after accounting for demographic variables and observed avoidance in a close friendship at 13.

Table 7.

Predicting Observed Avoidance in Romantic Relationships (Age 24))

	β	Δ R2	Total R2
Step 1.
Gender (1=M; 2=F)	−.24**
Family Income	−.18
			.128*
Step 2.
Observed Avoidance (13)	.14
		.021	.149*
Step 3.
Number of Boyfriends/Girlfriends (13)	.30**
		.120***	.269**

Note.

^*p <.05.

^**p ≤.01.

^***p ≤.001. βs are from final model.

Hypothesis 4: Conflictual friendships from ages 14–16 will mediate the relationships between romantic churning at 13 and later romantic relationship outcomes.

Using bootstrapped confidence intervals, all potential indirect effects were examined. There was a significant indirect effect from total boyfriends/girlfriends by 13 to abuse at 18 through total conflict/betrayal from ages 14–16 (β=.14, 95% CI =.021, .249). The effect from the number of boyfriends/girlfriends to abuse/coercion was no longer significant when close friendship conflict was included (β =.13, p=.36). This suggests that the effect of a greater number of boyfriends/girlfriends by 13 on abuse/coercion at 18 was explained by conflictual friendships from 14–16. No other significant indirect effects were found.

Hypothesis 5: Romantic churning will predict problematic relationship outcomes over and above baseline indicators of social functioning.

Next, to examine the possibility that qualities of the adolescent might underlie both having a greater number of boyfriends/girlfriends and later conflict, hostility, avoidance and closeness, we examined attachment security, social competence, and friendship quality as reported by the target adolescent at 13 as covariates. We found that attachment security (r=−.21, p=.01) and social competence (r=−.26, p=.001) were both significantly negatively correlated with the number of boyfriends/girlfriends reported at age 13. In addition, attachment security was significantly positively associated with peer preference at age 16 (r=.41, p=.001), and negatively associated with conflict and betrayal in close friendships from ages 14–16 (r=−.27, p=.001). However, the number of boyfriends/girlfriends remained a significant predictor when these variables were included, with the exception of the prediction to observed hostility at age 18, which remained marginally significant when social competence was included as a covariate (β=.27, p=.07). Overall, results suggest that the effect of romantic churning in early adolescence on later relationship conflict, hostility, and avoidance is not better explained by attachment security, baseline social competence, or friendship quality.

Post-hoc analyses

To determine whether target adolescents or their partners might be driving the effects of our dyadic findings, we examined predictions to each partner separately. The number of boyfriends/girlfriends at age 13 predicted conflict and betrayal for both the target participants (β=.20, p=.005) and their close friends (β=.32, p=.001). In terms of observed hostility with romantic partners at age 18, the number of boyfriends/girlfriends only significantly predicted observed partner hostility (β=.38, p=.003). In terms of abuse/coercion with romantic partners at age 18, the number of boyfriends/girlfriends predicted abuse/coercion for both participants (β=.27, p=.01) and partners (β=.39, p=.001). In terms of observed avoidance at age 24, the number of boyfriends/girlfriends predicted observed avoidance for both participants (β=.35, p=.001) and romantic partners (β=.21, p=.05). With the exception of observed participant hostility, these results suggest that the number of boyfriends/girlfriends by age 13 predicts both participant and partner behavior and perceptions of the relationship.

Discussion

Early adolescent romantic interests are often considered trivial and short-lived by adults who observe and interact with adolescents. The results of this study suggest that though they may be short-lived, for at least some adolescents, these early experiences may be far from trivial in their implications. This study identified a problematic developmental track that is marked by a rapid “churn” of romantic interests early in adolescence, potentially at the expense of developing same-gender friendships, emotion regulation skills, and conflict resolution capabilities. After accounting for gender, family income, social competence, attachment security, and qualities of early adolescent friendships, those adolescents who reported having had more boyfriends or girlfriends by age 13 appeared to develop more problematic peer relationships by mid-adolescence and more problematic romantic relationships by both late adolescence and early adulthood. These unique, long-term, multimethod findings build on prior research, which suggests that romantic involvement in early adolescence, before adolescents have developed necessary emotion regulation skills, is associated with depression (Compian, Gowen, & Hayward, 2004; Davila et al., 2009) and externalizing problems (Connolly et al., 2013) and other research that suggests romantic relationship dissolutions are associated with increases in anger and sadness (Joyner & Udry, 2000; Monroe et al., 1999; Sbarra, 2006; Slotter, Gardner, & Finkel, 2010). In addition, these findings are consistent with developmental theory that suggests that early adolescence is normatively marked by few, fledgling romantic interests, while adolescents focus on personal identity development and platonic peer relationships (Brown, 1999; Furman & Shaffer, 2003). These findings suggest that beginning adolescence in an off-track way by churning through multiple romantic interests may have long-term effects on development, likely as adolescents miss out on key skills typically formed in same-gender friendships and instead develop a potentially compulsive approach to relationships.

First, although the definition of a boyfriend or girlfriend may vary for early adolescents (Collins, Welsh, & Furman, 2009), our data suggests that a larger number of boyfriends or girlfriends was, in fact, related to other indicators of precocious romantic involvement. These data provide support to the idea that having a large number of boyfriends or girlfriends by age 13 represents a developmentally off-track course. By mid-adolescence, individuals who had reported a larger number of boyfriends or girlfriends at age 13 were rated as less preferable as companions within the broader peer group. A possible explanation is that such adolescents were distracted from investing in same-gender peer relationships that are typically primary at this age, and, as a result, they became decreasingly desirable as companions (Laursen & Williams, 1997; Zimmer-Gembeck, 1999). Alternatively, such adolescents’ social lives may have been “high drama,” which could become less appealing to their peers as they mature (Allen et al., 2014). It is important to note that this finding reflects a relative change in peer preference (i.e., predicting future peer preference even after accounting for baseline levels of peer preference). Although adolescents who churned through romantic interests at 13 were equally desirable as peer companions as other adolescents initially, they became less desirable over time, likely reflecting neglected opportunities to hone relational skills. In addition, they had friendships that were increasingly characterized by conflict and betrayal, according to both participants and their close friends. It appears likely that these adolescents’ social lives were filled with drama and frequent relationship dissolutions. Such adolescents may have devalued relationships (and likely treated friendships as less important than romantic relationships; Laursen & Williams, 1997; Zimmer-Gembeck, 1999), which would then contribute to hostility and mistrust within the friendship (although we note that our data cannot speak to this causal explanation).

Adolescents who churned through romantic interests also formed observably hostile romantic relationships by age 18. These adolescents may have developed (or already held) schemas of relationships as transitory and marked by conflict. Such expectations seem likely to elicit hostility within relationships and from both partners (Downey et al., 1998). In addition to eliciting hostility from partners, adolescents who were repeatedly focused on romantic interests early in adolescence potentially failed to learn key social skills during adolescence, and such lack of skill is associated with poor coping, including aggression (Mulford & Giordano, 2008). Both participants and their respective romantic partners reported perpetrating more abuse and coercion towards each other by age 18. These findings indicate that romantic churning in early adolescence has the potential to explain not just the adolescent’s own future behaviors, but also the behavior of those they choose as partners, possibly due to either a selection effect or a self-fulfilling prophecy effect (Loeb, Hessel, & Allen, 2016). In addition, conflictual friendships from 14–16 mediated the relationship between romantic churning at 13 and abuse/coercion at age 18. This is consistent with the idea of a developmental cascade and with prior research that found problematic peer relationships in adolescence predicted coercion in romantic relationships in adulthood (Ha, Otten, McGill, & Dishion, 2019). It appears that beginning adolescence on a risky track characterized by romantic churning may cascade forward into problems in different types of close relationships throughout adolescence.

By age 24, participants who pursued multiple romantic interests by age 13 had romantic relationships characterized by avoidance during conflict. One way to view this is as the culmination of a process in which participants experience relationships as increasingly unsatisfying and dysfunctional (first within the broader peer group, and later in their experience of hostile romantic relationships in late adolescence). Findings suggest a developmental trajectory of romantic relationships of adolescents who churn through romantic interests early in adolescence. Specifically, rather than continuing to be conflict-ridden, romantic relationships appear to become characterized by an absence of positivity, perhaps as individuals, over time, give up on the attempt to navigate conflict and select similarly avoidant partners. It is particularly notable that this avoidant pattern was displayed by individuals who, as early adolescents, appeared to value and take great interest in engaging in many romantic relationships.

An alternative view of such romantic churning is that individuals who report a large number of romantic interests by 13 are a select group from their peers in some important respects. The current study examined several competing explanations for the primary findings by exploring the role of family income, gender, attachment style, social competence, and baseline friendship quality as potentially driving both early romantic behavior and later outcomes. It was found that, although social competence and attachment security at 13 were associated with fewer boyfriends/girlfriends, neither these, nor any other factors examined significantly altered findings when added to predictive models. The one exception is that the link between the number of boyfriends/girlfriends and observed hostility at age 18 became nonsignificant when social competence was included as covariate. Overall, these findings provide more support for the idea that a rapid churn of romantic interests itself has a negative influence, rather than just reflecting an underlying social deficit (perhaps with the exception of the development of observed hostility) within the adolescent. Such romantic churning seems to warrant further attention from researchers and others concerned with adolescent development.

There are some important limitations to note. Given its long-term longitudinal nature, this study of necessity examined individuals who were adolescents in the late 1990s and early 2000s. Although this allowed us to examine long-term predictions, adolescent dating patterns have likely changed with the advent of social media and near-constant access to cell phones for a large number of adolescents. Future studies should build on these findings by examining how current dating patterns are linked to long-term outcomes for adolescents. In addition, while our data on dating/romantic behaviors provides some evidence that adolescents who report a larger number of boyfriends or girlfriends are engaging in more precocious dating behavior, adolescents may interpret “boyfriend/girlfriend” in a number of different ways and these relationships may have varied greatly in levels of intensity and duration. Although we accounted for several potential third variables that could explain both churning and the outcomes of interest, there are many other possible third variables that may underlie churning and negative outcomes. In addition, because this study did not have access to data prior to age 13, important questions remain about the underpinnings of romantic churning in early adolescence. It is likely that such churning may have its roots in early familial relationships, such as exposure to abuse, neglect, or lax parenting. Future studies should further examine the conditions in childhood that are related to romantic churning in early adolescence. In addition, although the longitudinal nature of the data is informative and allows us to examine relative change, it is nevertheless not sufficient to draw causal conclusions. We also did not have access to information on whether the reported romantic relationships were reciprocal, and we know that adolescents with reciprocal relationships tend to be better adjusted than those with nonreciprocal relationships (Carlson & Rose, 2007). We also did not have information about pubertal status or timing. Although consistent with other intensive observational studies of romantic dyads (Ha, Dishion, Overbeek, Burk, & Engels, 2014), our sample size became small for the romantic partner collection points. Finally, at age 13, participants did not specify whether their romantic interests were with opposite or same sex individuals. We also did not collect data on gender identity at age 13. Future research should examine potential differences in romantic churning for LGBTQ youth.