Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Jan 11.
Published in final edited form as: Emerg Adulthood. 2017 Mar 29;5(4):241–250. doi: 10.1177/2167696817699750

“Stay or Leave”: Predictors of Relationship Dissolution in Emerging Adulthood

Ann Lantagne 1, Wyndol Furman 1, Jamie Novak 1
PMCID: PMC6329602  NIHMSID: NIHMS978065  PMID: 30637181

Over the course of emerging adulthood, individuals gradually become more independent, assume more adult responsibilities, make decisions about their careers, and establish intimate relationships (Arnett, 2004). These years are characterized as a time of exploration and uncertainty with regard to several life tasks, including establishing romantic relationships. Across emerging adulthood, individuals may engage in a variety of romantic relationships on the path to determining what they want in a partner (Shulman & Connolly, 2013). Many will experience the dissolution of a number of romantic relationships; in fact, over a third of emerging and young adults have experienced a breakup within the past 20 months (Rhoades, Kamp-Dush, Atkins, Stanley, & Markman, 2011). Accordingly, one of the key developmental tasks of emerging adulthood is not only learning how to initiate and maintain a romantic relationship, but also learning when to end a relationship (Halpern-Meekin, Manning, Giordano, & Longmore, 2012).

Given the salience of breakups during emerging adulthood, it is important to understand why some relationships dissolve whereas other relationships are maintained. The present study examined the role of contextual, relationship, and individual predictors of relationship dissolution during emerging adulthood. We first explored which factors are relevant for predicting breakups in the short-term, or within the next 12 months after data collection. For those relationships that did not dissolve in the next 12 months, we then explored which factors are relevant for predicting breakups over the long-term, or within the 12 to 72 months following data collection. In other words, we examined which factors predicted the eventual dissolution of these relationships. As such, the present study determined whether predictors of relationship dissolution could be identified well in advance of an actual breakup, as well as shortly before a breakup.

Relationship Dissolution in Emerging Adulthood

A recent meta-analysis identified a number of predictors of nonmarital relationship dissolution (Le, Dove, Korn, Agnew, & Matsu, 2010). Consistent with current models of marital dissolution (Halford, Markman, Kline, & Stanley, 2003), such predictors can be categorized into three levels: the contextual level, the relationship level, and the individual level.

Contextual variables describe the larger environment in which the relationship exists, and can include variables such as living in a stressful environment or experiencing stressful life events. To the best of our knowledge, there is no existing research linking stressors to the dissolution of emerging adults’ relationships. However, those individuals who find emerging adulthood to be particularly stressful may have fewer resources for handling the demands of a romantic relationship (Shulman & Connolly, 2013). Moreover, stressors are negatively associated with the quality of emerging adults’ relationships (Kogan et al., 2013), suggesting that stressors may relate to dissolution.

At the relationship level, relationship qualities are fundamental in predicting dissolution. The qualities of romantic relationships are particularly important in emerging adulthood as they are linked to concurrent adjustment (Collibee & Furman, 2015). In terms of positive qualities, lower levels of support and validation have been associated with an increased likelihood of the relationship dissolving (Gable, Gonzaga, & Strachman, 2006). High levels of conflict have also been consistently identified as a primary reason for relationship dissolution in emerging adulthood (Dailey, Rosetto, Pfiester, & Surra, 2009; Surra & Longstreth, 1990).

Individual characteristics such as psychopathology also predict nonmarital relationship dissolution. Psychopathology is characterized by interpersonal difficulties, and individuals with higher levels of psychopathology have greater relationship distress (Whisman & Beach, 2001). Emerging adulthood is a particularly salient time to examine psychopathology, as individuals have an increased risk of mood disorders and substance use during this time (Kessler et al., 2005). In fact, internalizing symptoms are associated with relationship dissolution in dating relationships (Khaddouma, Norona, & Whitton, 2015). Little research has examined how other forms of psychopathology such as externalizing symptoms or substance use are related to nonmarital dissolution, although such factors are predictive of divorce (Newcomb, 1994).

Also at the individual level, it is informative to examine the role of romantic appeal, or individuals’ beliefs that they are attractive partners and dating the people they would like to be dating (Harter, 2012). Individuals who do not believe they are attractive or who are not dating the people they would like to date are likely to be involved in relationships that dissolve more quickly. Romantic appeal is particularly informative to examine during emerging adulthood, when establishing a romantic relationship becomes a salient developmental task (Masten, Burt, & Coatsworth, 2006).

Taken together, the existing literature provides some information about the nature of relationship dissolution in emerging adulthood, yet notable limitations exist. Although studies have examined contextual, relationship, or individual predictors, relatively little work has simultaneously examined all three levels (for exceptions, see Felmlee, Sprecher, & Bassin, 1990 & Khaddouma et al., 2015). Studies examining all three levels are important in order to obtain a more comprehensive picture of the predictors of dissolution and the interplay among them.

Furthermore, research has typically examined relationship status as a dichotomous outcome at a specific point in time in the relatively immediate future– i.e., several months later, has the relationship dissolved or not? However, significant variation exists in how much longer a relationship exists before dissolving. Assessing the length of time until dissolution yields a more sensitive index than simply looking at the dichotomous outcome of whether a relationship dissolved or not at some later point in time (MacCallum, Zhang, Preacher, & Rucker, 2002).

A final related limitation is that studies of emerging adults’ relationships have predominantly examined if we can predict whether a breakup will occur within the immediate future, such as the next few months or perhaps the next year or two at most. Only a few studies of emerging adults’ romantic relationships have examined relationship dissolution over the long term as well (e.g., DeMaris, 2000; Sprecher, 1987). It would be informative to follow a relationship over multiple years to determine if there are indications of relationship dissolution long before a breakup actually occurs.

The Present Study

The present study examined contextual, relationship, and individual predictors of relationship dissolution in emerging adulthood. The contextual level included the cumulative number of stressful life experiences from the past year. The relationship level included support and conflict. The individual level included psychopathology (internalizing symptoms, externalizing symptoms, & substance use), as well as romantic appeal. We examined these variables individually and then together across all three levels in order to better understand which factors contribute to relationship dissolution in emerging adulthood.

As we were interested in the role of relationship qualities, we gathered data from ongoing relationships. Because they were ongoing, relationships varied in length at the time we collected data on them. Consequently, we framed our questions about dissolution in terms of what variables predicted the number of months between data collection and relationship dissolution, rather than in terms of what variables predicted the overall length of the relationship. Due to variations in the length of participants’ relationships at data collection, we included relationship length at the time of data collection as a control variable.

Contextually, we expected that higher levels of stressful life events would be associated with relationships dissolving sooner after the point of data collection. At the relationship level, we anticipated that lower levels of support and higher levels of negative interactions would also be related to relationships dissolving sooner. At the individual level, we expected higher levels of internalizing symptoms, externalizing symptoms, and substance use, as well as lower romantic appeal would be related to relationships dissolving sooner.

In order to identify the predictors of relationship dissolution in both the short term and long term, we tracked the course of relationships for a total of six years after data on predictors were collected. We were first interested in which variables predicted relationship dissolution in the next year after data collection-- i.e. in the next twelve months, how many more months until the relationship dissolved, or if it lasted for another year. For those relationships that did not dissolve within the next year, we then examined which variables predicted how much longer until the relationship dissolved, or if the relationship still existed after five additional years. We did not make separate hypotheses regarding predictors of dissolution in the next year after data collection versus over the long-term, as we had no empirical basis for postulating how the predictors would differ. Predictors differ for shorter and longer time to dissolution in adolescent romantic relationships (Appel & Shulman, 2015), and predictors also vary for relatively more immediate versus eventual divorce (Gottman & Levenson, 2000). No existing studies have considered how such predictors might differ in emerging adults’ romantic relationships. However, it is informative to know if these predictors are the same or if they differ, as the latter would suggest that different processes might be at play.

Method

Participants

Data were drawn from a larger longitudinal study focusing on close relationships and psychosocial adjustment across adolescence and young adulthood. Two hundred tenth grade participants (100 males, 100 females) were recruited from a Western metropolitan area. To obtain a diverse sample, we distributed brochures and sent letters to families residing in various zip codes and to students enrolled in schools in ethnically diverse neighborhoods. Because letters were sent to many families who did not have a 10th grader, we were unable to determine the participation rate. We contacted interested families with the goal of selecting a quota sample with equal males and females, and a racial/ethnic group distribution that approximated that of the United States. Families were compensated $25 to hear a description of the project; 85.5% expressed interest and carried through with the first assessment.

The sample consisted of 11.5% African Americans, 12.5% Hispanics, 1.5% Native Americans, 1% Asian Americans, 4% biracial, and 69.5% White, non-Hispanics. The sample was of average intelligence (WISC-III vocabulary score M = 9.8, SD = 2.44), and closely approximated national norms on various measures of psychosocial adjustment and substance use (see Furman, Low, and Ho, 2009). 55.4% of their mothers had a college degree, indicating that the sample was predominately middle or upper middle class. With regard to family structure, 57.5% resided with two biological or adoptive parents, 11.5% resided with a biological or adoptive parent and a stepparent/partner, and 31% resided with a single parent or relative.

In terms of sexual orientation, 89.3% reported that they were heterosexual at Wave 7 whereas the other participants said they were bisexual (4.2%), gay (2.4%), lesbian (2.4%), or questioning (1.8%). We retained the sexual minorities in the sample to be inclusive and because we had no theoretical or empirical basis for believing that the results would differ by sexual orientation.

Procedure

For the present study, predictor data were drawn from Waves 4 through 7 of the study when questionnaire and interview data were collected every 18 months (Wave 4 mean age = 19 years, .36 months, SD = 6.72 months; Wave 7 mean age = 23 years, 8 months). For relationships that had not dissolved by Wave 7, we obtained information on when that relationship dissolved from Waves 8 – 11. Participant retention was excellent (Wave 4: N= 195; Wave 5: N= 186; Wave 6: N= 185; Wave 7: N= 179). Those who participated in the study in Wave 7 did not differ from those who did not in terms of age, ethnicity, maternal education, or scores on the relationship quality variables in Wave 1.

Measures

Relationship length and length of time until dissolution

At each wave of data collection, participants were interviewed about their romantic relationships. In the beginning of the interview, they identified their most important relationship in the last year that had lasted at least a month. To avoid problems of retrospective reporting, we excluded past relationships. Furthermore, because we were interested in nonmarital romantic relationships in emerging adulthood, we removed marital relationships from analyses (N=14 relationships).

The participants also reported on the length of the relationship at the time of data collection. The length of time until dissolution was determined from the interview report in subsequent waves (see Results section for more detail).

Life Stress

Participants completed the Life Events Questionnaire- Adolescent Form (LEQ-A; Masten, Neeman, & Andenas, 1994) in Wave 4 and the Life Events Questionnaire-Young Adult Form (LEQ-YA) in Waves 5–7 (Masten & Tellegen, 2012). Participants were asked to indicate which of a variety of potentially stressful life events they had experienced in the past year (e.g., “I became seriously ill or was injured”). Participants’ scores on this measure were the total number of negative life events they endorsed among the 58 life events that were common to the LEQ-A and LEQ-YA.

Network of Relationships Inventory: Behavioral Systems Version (NRI; Furman & Buhrmester, 2009)

The short version of the NRI included five items regarding social support in their relationship, and six items regarding conflict and negative interactions such as antagonism and criticism. Using a five-point scale, participants rated how much each item characterized their romantic relationship. Support and negative interaction scores were calculated by averaging the relevant items (M α = .89 & M α = .92, respectively).

Internalizing Symptoms

Internalizing symptoms were assessed using the trait scale of Spielberger’s (1983) State-Trait Anxiety Inventory (STAI; M α = .92), the Beck Depression Inventory (BDI; Beck, Rush, Shaw, & Emery, 1979; M α = .86), and an abbreviated 20-item version of the internalizing scale from the Adult Self Report (Achenbach, 1997). These three scales were substantially correlated with one another (M r = .66) and thus, combined into a composite. The derivation of this composite and other composites subsequently described involved a number of steps. The various measures used to create this composite had different numbers of points on their scales, which means the scores on the different scales are not comparable. Accordingly, we standardized scale scores across all waves to render the scales comparable with one another, a recommended procedure that retains differences in means and variance across age, and does not change the shape of the distribution or the associations among the variables (Little, 2013). Standardized scores on the three measures were then averaged to form the internalizing symptoms composite.

Externalizing Symptoms

Participants completed an abbreviated 26-item version of the externalizing scale on Achenbach’s (1997) Adult Self-Report (M α = .87). Friends and mothers reported on the participant’s externalizing symptoms by completing abbreviated 19 item versions of Achenbach’s (1997) Adult Behavior Checklist (M α’s = .84 & .89). These three scales were substantially correlated with one another (M r = .66); and were standardized and averaged to form the externalizing composite.

Substance Use

Participants completed the Drug Involvement Scale for Adolescence (Eggert, Herting, & Thompson, 1996). Questionnaires were administered by computer-assisted self-interviewing techniques to increase candor. Participants’ use of beer, wine, liquor, marijuana, and other drugs over the last 30 days was assessed. Frequency of substance use was scored on a 7-point scale ranging from never to every day. Participants also completed a 9-item measure assessing adverse consequences from substance use (M α = .92) and a 16-item measure assessing difficulties in controlling substance use (M α = .90). Likewise, as part of their version of the Adolescent Self-Perception Profile (Harter, 1988), friends answered four questions about the participant’s alcohol and drug use and problems related to substance use. Items were averaged for the friend report of the participant’s substance use and problems (M α = .82).

To obtain a measure of alcohol use, we averaged the participants’ reports of drinking beer, wine, and/or liquor. To measure drug use, we averaged participants’ reports of marijuana use and other drug use. Participants’ reports of adverse consequences and control problems were averaged to measure problem usage. Participants’ different reports and friends’ reports were then averaged to obtain a composite measure of substance use.

Romantic Appeal

Participants, friends, and mothers rated the participant’s romantic appeal using an abbreviated form of Messer and Harter’s (1986) scale on the Adult Self-Perception Profile (M α = .81). This scale consisted of five items using a four-point structured alternative format. The three reporters’ perceptions of romantic appeal were significantly related to each other (participant-mother M r = .55; participant-friend M r = .50, friend- mother M r = .42, all ps < .001). Accordingly, we computed composite measures of appeal by standardizing and averaging the different reporters’ perceptions.

Results

Preparation of Data

We collected data on variables such as relationship qualities while a relationship was ongoing. Consequently, we framed our questions about dissolution in terms of how many months passed from data collection to relationship dissolution, rather than in terms of the total number of months the relationship lasted. To calculate the time until dissolution, the length of the relationship at the time when predictor data were collected was subtracted from the total length of the relationship. For example, if we collected data on a six-month relationship and the participant subsequently reported that the relationship dissolved after a total length of fourteen months, the time until dissolution was eight more months. The relationships varied in length when the data were collected so we controlled for the pre-existing length of the relationship in our analyses.

Multilevel modeling allowed us to include multiple waves of data on the same relationship as well as data on multiple relationships. Estimations of the length of the relationship and the time until dissolution were calculated separately for each point of data collection on a relationship. The median length of the romantic relationship at data collection was 13 months and the median time to dissolution was 18 months. 79.4% of the time, the relationship had dissolved within 72 months after data collection.

Predictors of Relationship Dissolution in the next 12 Months

First, we examined the predictors of relationship dissolution within the next 12 months. We were interested in which variables predicted the number of months until the relationship dissolved. If the relationship had not dissolved within the next 12 months after data collection, we capped the score at 12 months because we were interested in the predictors of dissolution in the short-term and did not want these analyses to be influenced by variations in relationships that only dissolved in the long term. For example, a relationship that dissolved 18 months after data collection, and another one that dissolved 60 months after data collection would receive the same score of 12 months in the analyses of dissolution in the short-term; however, they received their original scores in the analyses of long-term dissolution described subsequently.

To test hypotheses, a series of multilevel models were conducted using the statistical program Hierarchical Linear Modeling (HLM Version 6.0; Raudenbush, Bryk, & Congdon, 2004). HLM takes into account the nested nature of the data in a longitudinal study and permits missing data by using full information maximum likelihood (FIML) to estimate parameters. Models had the following form:

Level 1: Yti = β0i + β1i (An individual/relationship/contextual predictor) + β2i (Relationship length at time of data collection)+ β3i (Age) + rti
Level 2: β0i = γ00 + γ01 (gender) + u0i
β1i = γ10
β2i = γ20
β3i = γ30
Open in a new tab

In these models, Yti represented the number of months until the relationship dissolved at time t for individual i with the maximum score of 12 months. Age and relationship length at the time of data collection were included as control variables. Data on relationships over the next 12 months were available on 326 relationship observations from 158 participants.

Analyses of each predictor

We first examined each predictor variable one by one to ascertain whether it was associated with the time until dissolution within the next 12 months after data collection (see Table 1 for results). At the relationship level, lower levels of support were predictive of the relationship dissolving sooner. Lower romantic appeal was predictive of the relationship dissolving sooner. Relationship length at the time of data collection was also associated with the amount of time until dissolution, such that shorter relationships tended to dissolve sooner.

Table 1.

Multilevel Models Testing the Associations between Contextual, Relationship, or Individual Level Predictors and the Time until Dissolution in the Short-Term

Stressful Life Events Relationship Support Negative Interactions Romantic Appeal Internalizing Symptoms Externalizing Symptoms Substance Use
Intercept (β0i) 5.69(3.11) 3.83(3.09) 0.85(0.51) 4.98(2.94) 4.85(2.97) 4.38(3.00) 5.03(3.00)
Age (β2i) 0.13(0.15) 0.04(0.14) 0.06(0.14) 0.14(0.14) 0.15(0.14) 0.17(0.14) 0.15(0.14)
Relationship Length at Data Collection (β3i) 0.05**(0.01) 0.04**(0.01) 0.05**(0.01) 0.04**(0.01) 0.05**(0.01) 0.05***(0.01) 0.05**(0.01)
Predictor (β1i) −5.63(4.55) 0.92**(0.30) −0.11(0.34) 0.68*(0.31) −0.32(0.25) 0.11(0.31) −0.36(0.37)
Gender Main Effect (γ01) 0.84†(0.50) 0.65(0.50) 0.67(0.51) 0.77(0.50) 0.97†(0.51) 0.83(0.51) 0.72(0.51)
Open in a new tab

Notes.

†

p < .10;

*

p < .05;

**

p < .01;

***

p < .001.

This table presents the results when each of the contextual, relationship, and individual variables were entered individually. The primary numbers in the table are the unstandardized coefficients for the fixed effects. A negative beta would indicate that higher scores on the predictor variable were associated with sooner dissolution. Standard errors are in parentheses.

Life stress, negative interactions, internalizing symptoms, externalizing symptoms, and substance use were not predictive of the time until dissolution in the short term. Age and gender were also not significant in these models.

Analyses of all significant predictors

Next, we simultaneously examined all of the variables that had been found to be predictive in our first set of analyses to determine which variables made a unique contribution in predicting the time until dissolution in the short-term. We used the same model as in the first set of analyses except that we included both significant variables (relationship support & romantic appeal). Only relationship support uniquely predicted the time until dissolution within the next 12 months, B = 0.86, t(277) = 2.59, p < .05.

Dichotomous analyses

Typically investigators have simply asked whether a relationship dissolved or not at a specific point in time. We chose to examine the amount of time until dissolution because statistically, a continuous measure is a more sensitive index than a dichotomous measure (MacCallum, et al., 2002). However, we also conducted supplementary analyses with a dichotomous measure of relationship dissolution. Specifically, we conducted hierarchical Bernoulli regression models, which are analogous to logistic regression models (Raudenbusch & Bryk, 2002). The models were the same as the first set of analyses except that the outcome variable was dichotomous. We coded our outcome so that our dichotomous analyses would be consistent with the direction of our continuous analyses (0= dissolved within the next 12 months; 1 = did not dissolve within the 12 months following data collection). Predictors were examined one by one. Within the next 12 months following data collection, 36.7% of the relationships dissolved. Relationship support was again a significant predictor (B = 0.39, t(280) = 2.45, p < .05), and romantic appeal was now a near significant trend (B = 0.31, t(293) = 1.94, p = .053).

Predictors of Dissolution 12 to 72 Months After Data Collection

If the relationship did not dissolve within the next year, we were then interested in what was predictive of the time until dissolution over a longer time span—i.e. between 12 and 72 months after data collection. We removed relationships that had dissolved within the next 12 months so that these analyses would not overlap with the prior analyses. In the event that the relationship had not dissolved within 72 months after the predictor variables were collected, we capped the total length of time until dissolution at 72 months. This length was chosen because we were able to determine whether all relationships lasted for at least 72 months if the participant was still in the study in our most recent wave of data collection. In six instances, the participant dropped out of the study; these relationships were deleted from these analyses, as we were unable to determine when they dissolved. Data were available on 190 relationship observations from 107 participants.

Analyses of each predictor

We first examined each predictor variable one by one to determine whether it was associated with the time until dissolution in the long term, or an additional five-year period (see Table 2). At the contextual level, higher numbers of stressful life events were predictive of the relationship dissolving sooner. At the relationship level, lower levels of support and higher levels of negative interaction were predictive of relationships dissolving sooner. At the individual level, individuals with lower romantic appeal had relationships that dissolved sooner. Higher levels of externalizing symptoms and higher levels of substance use were predictive of relationships dissolving sooner. Younger individuals’ relationships also dissolved sooner. Finally, relationships that were longer at the time of data collection dissolved sooner. This finding is to be expected as the more time a particular relationship had already existed, the less time it had before reaching the time-point when that relationship dissolved; thus, this finding illustrates the importance of controlling for prior length, but it is not a substantive finding.

Table 2.

Multilevel Models Testing the Associations between Contextual, Relationship, or Individual Level Predictors and the Time until Dissolution in the Long Term

Stressful Life Events Relationship Support Negative Interactions Romantic Appeal Internalizing Symptoms Externalizing Symptoms Substance Use
Intercept (β0i) 9.15(18.57) −22.35(17.96) −0.77(17.42) −3.20(19.29) 0.39(18.10) −4.78(17.85) −20.91(17.11)
Age (β2i) 2.25*(0.87) 2.20*(0.85) 3.03**(0.82) 2.36**(0.83) 2.36**(0.88) 2.62**(0.87) 3.41**(0.84)
Relationship Length at Data Collection (β3i) −0.49***(0.08) −0.46***(0.08) −0.52***(0.08) −0.47***(0.08) −0.49***(0.08) −0.51***(0.08) −0.51***(0.08)
Predictor (β1i) −80.07**(30.33) 6.25**(1.92) −7.00**(2.18) 7.72***(1.90) −3.52†(1.89) −4.63*(2.23) −5.17*(2.50)
Gender Main Effect (γ01) 1.80(5.02) 0.17(4.96) −0.33(5.17) 0.03(4.94) 3.34(5.08) 0.33(5.11) 0.42(5.21)
Open in a new tab

Notes.

†

p < .10;

*

p < .05;

**

p < .01;

***

p < .001.

This table presents the results when each of the contextual, relationship, and individual variables were entered individually. The primary numbers in the table are the unstandardized coefficients for the fixed effects. A negative beta would indicate that higher scores on the predictor variable were associated with sooner dissolution. Standard errors are in parentheses.

Analyses of all significant predictors

Next, we simultaneously examined all of the variables that had been found to be predictive of the time until dissolution in the long term to determine which variables made a unique contribution. We used the same model as in the set of analyses where we entered each predictor individually, except that we included all six variables that were significantly predictive when examined one by one (stressful life events, relationship support, negative interactions, externalizing symptoms, substance use, & romantic appeal). When examined simultaneously, the number of stressful life events, negative interactions, and romantic appeal each uniquely predicted the time until dissolution over the additional five-year period (B’s = −64.52, −4.55, & 5.25, respectively, ps < .05).

Dichotomous analyses

Finally, we conducted supplementary analyses to determine which variables predicted whether a relationship had dissolved or not over the additional five year period. We conducted hierarchical Bernoulli regression models with a dichotomous outcome (0 = dissolved during the additional 5-year period, 1 = did not dissolve over the additional 5 years). Predictors were examined one by one. Sixty-seven percent of the relationships dissolved sometime in the additional five-year period. Support, negative interactions, and romantic appeal continued to be significantly predictive of relationship dissolution (B = 0.56, t(179) = 2.10, p < .05; B = −0.76, t(180) = −2.45, p < .05; B = 0.70, t(183) = 2.60, p < .05). Higher levels of internalizing symptoms were also significantly predictive of relationship dissolution (B = −0.45, t(182) = −2.08, p < .05). Stressful life events became a nonsignificant trend (p = .08). Age, externalizing symptoms, and substance use were now nonsignificant (all ps > .10). Thus, results were similar to those with the continuous outcome of relationship dissolution, except somewhat weaker, as would be expected with a dichotomous measure.

Discussion

The present study highlights the importance of understanding why romantic relationships end during emerging adulthood, a time when establishing and maintaining an intimate relationship becomes a salient developmental task (Masten, Burt, & Coatsworth, 2006; Shulman & Connolly, 2013). Findings demonstrate that during emerging adulthood, relationship and individual factors are predictive of relationship dissolution within the next year after data collection, whereas contextual, relationship, and individual factors are predictive of relationship dissolution over the five following years. In addition to contributing to our understanding of short-term predictors of dissolution, findings also underscore the merits of examining predictors of relationship dissolution over a longer term. Notably, we were able to predict well in advance which relationships were likely to dissolve and which would last.

Predictors of the Amount of Time Until Dissolution

The present study examined multiple levels of predictors of time until dissolution. At the contextual level, higher numbers of stressful life events were linked to relationships dissolving sooner. Experiencing elevated levels of stress may deplete the cognitive resources an individual is able to allocate to sustaining a relationship (Neff & Karney, 2009). This is particularly salient during a developmental period in which individuals are already tasked with handling changes across multiple domains, including work, relationships, and individual goals (Ranta, Dietrich, & Salmela- Aro, 2013). Additionally, many emerging adults currently find themselves in particularly uncertain financial conditions (Shulman & Connolly, 2013). Experiencing elevated levels of stress outside of the relationship can spill over into the romantic relationship (Bolger, DeLongis, Kessler, & Wethington, 1989). Higher numbers of stressors may cause an individual to engage in more negative interactions with their partner, or to withdraw from the relationship and spend less time with their partner (Neff & Karney, 2009).

Consistent with previous research (see Le et al., 2010), the present study also found that relationship qualities were important predictors of relationship dissolution. Lower levels of support predicted dissolution in both the short- and long-term. An insufficient amount of support may be particularly relevant during emerging adulthood, when an important goal is having a high level of intimacy in a relationship, as reflected by open communication, interdependence and support (Zimmer-Gembeck, Hughes, Kelly, & Connolly, 2011). Emerging adults themselves report that they are more likely to end relationships that do not meet these goals (Norona, Olmstead, & Welsh, 2016). Many emerging adults are in and out of relationships across this time, switching between steady romantic relationships and brief romantic encounters (Arnett, 2004). Thus, the nature of relationships in this developmental period differ, and a lack of supportive interactions may be an indicator that the relationship is more exploratory in nature and thus, likely to end sooner.

Higher levels of negative interactions were also predictive of relationship dissolution over the long term. Antagonism, criticism, and conflict may accumulate across time and gradually lead to the deterioration of the relationship. Negative interactions that arise because of an individual’s characteristics, such as insensitivity, an unwillingness to compromise, or untrustworthiness, may also have a long-term impact on a relationship because they may be difficult to change. Finally, emerging adults are faced with coordinating multiple life tasks, such as their careers and their romantic relationships (Shulman & Connolly, 2013). Individuals who have higher levels of conflict in their relationships may be having difficulty balancing the demands of these different domains.

Taken together, the associations between relationship qualities and dissolution may stem from the fact that emerging adulthood is a time of exploring what one wants in a relationship (Arnett, 2000). Individuals may engage in a variety of relationships across this time and begin to determine which characteristics are desirable in a long-term partner. Some may learn that a certain relationship is not supportive or that it does not fulfill their needs, and dissolve that relationship.

At the individual level, lower levels of romantic appeal were predictive of relationships dissolving sooner. Perhaps individuals who are not confident in their romantic appeal have been shaped by previous unsuccessful romantic experiences and thus find it more difficult to maintain their relationships. Additionally, they may not have achieved a sense of relational identity, a salient developmental task in emerging adulthood (Erikson, 1968). Lacking a strong sense of self in the romantic domain, they may find it difficult to integrate their needs with others in a relational context (Beyers & Seiffge-Krenke, 2010).

Finally, variables at contextual, relationship, and individual levels were all predictive of relationship dissolution. Support, conflict, and romantic appeal may reflect a common underlying element, such as romantic competence. Those who are lower in romantic competence may be less adept at maintaining relationships and more prone to having them dissolve (Shulman, Davila, & Shachar-Shapira, 2011).

Predictors of Dissolution in the Short- and Long-Term

Most previous studies have examined which relationships have dissolved within the relatively immediate future, such as the next few months or a year or so at most. Consequently these studies are only examining a minority of the relationships that dissolve during emerging adulthood (34.4% of relationships in the present study). A more complete understanding of dissolution requires identifying relationships that dissolve over the long-term as well as those that dissolve in the near future. One of the most striking findings is that we were able to predict relationship dissolution over a substantially longer period of time than has been typically looked at. In fact, we were able to predict the time until dissolution even when we did not include the relationships that dissolved within the next year. In effect, we were able to predict which relationships would dissolve several years later as well as those that would dissolve within the next year. Thus, we were able to determine that the factors that predicted dissolution existed long before the relationships dissolved.

Interestingly, life stress, negative interactions, substance use, and externalizing symptoms were significant predictors of relationship dissolution only over the long term. Indeed, it may take some time for these variables to erode away at the foundations of the relationship. Perhaps over the long term, these features make one’s romantic partner increasingly dissatisfied with their interpersonal exchanges and ultimately lead to the relationship’s demise.

On a related note, it is interesting that we did not identify any individual level predictors that singularly predicted dissolution in the short term. Such variables may be less likely to be indices of the ongoing relationship quality per se, and instead could be significant discrete events, such as acts of infidelity or aggression, that lead to a sudden transformation of the relationship. Clearly, further work is needed to understand why some predictors are significant as both short- and long-term predictors, whereas others only predict dissolution over the long term. It will be important to explore and identify other variables that might only be predictors in the short term as well.

Multiple Levels of Predictors

Very few existing studies have examined factors concurrently across all three levels of predictors (for exceptions, see Felmlee et al., 2000 & Khaddouma et al., 2013). The present study not only examined variables individually, but also examined their unique contributions when all of the significant predictors across the contextual, relationship, and individual levels were included.

Consistent with prior work (Le et al., 2010), support was uniquely predictive of the amount of time until dissolution in the next year. Although significant in the individual analyses of variables, romantic appeal did not uniquely contribute to the prediction of relationship dissolution over the short term. Perhaps this variable plays an indirect role by affecting how supportive or unsupportive the relationship is. For example, individuals with lower levels of romantic appeal may be less adept at initiating and sustaining supportive interactions in their relationship.

In terms of long-term predictors of dissolution across levels, stressful life events, negative interactions, and romantic appeal were all significant and unique predictors. Marital distress and dysfunction have long been hypothesized to emerge from the combination of stressful events, enduring vulnerabilities (e.g. psychopathology), and poor adaptive processes that may be reflected in interaction patterns within a relationship (Karney & Bradbury, 1995). The present findings are consistent with such a model, as each of these factors provided unique contributions in predicting nonmarital relationship dissolution over the long-term as well. Thus, the processes underlying relationship dissolution in married and unmarried couples may be similar in a number of respects. What is less evident is how these processes may differ between marriages and other romantic relationships. In particular, relationships of most emerging adults involve less dedication and fewer constraints than those in long-standing committed relationships such as marriages (Stanley & Markman, 1992), but we do not know how such distinctions may differently shape predictors.

Limitations and Future Directions

The present study provides a significant contribution by examining predictors of dissolution in emerging adulthood over both the short- and long-term. However, some limitations exist. The present study mainly utilized self-report data, occasionally supplemented by friend and mother reports. Future studies should incorporate other methods such as observational and interview data.

Additionally, we did not collect data on the romantic partners’ individual characteristics, their perceptions of the relationships, or the contextual factors influencing them. Romantic partners’ experiences should have as much influence on relationship dissolution as the participants’. Discrepancies in perceptions of variables such as closeness impact whether a relationship dissolves (Frost & Forrester, 2013).

One of the most interesting findings is that some of the factors that predicted how long a relationship would last before dissolving existed long before the relationships dissolved. However, what we do not know is how these factors led to the eventual dissolution of some relationships. It would be interesting to see how the predictor variables and the relationships change as they approach dissolution.

Finally, although the sample was comparable to national norms on a number of measures of adjustment, the participants were primarily from middle or upper-middle class families. Future studies need to explore the role that factors such as socioeconomic status play in relationship dissolution during this time.

Despite these limitations, the present study has important implications for understanding breakups during emerging adulthood. Furthermore, the fact that we were able to identify multiple predictors well in advance of when relationship dissolution actually occurred has important implications for clinical practice as well. Levels of stress, conflict, and romantic appeal may all be factors that can be preventatively targeted by clinical interventions to improve relationships during this time.

Taken together, the present study demonstrated that contextual, relationship, and individual factors all contribute to relationship dissolution during emerging adulthood. Moreover, the present study examined these factors over a total of six years, which is an extensive time span. By doing so, we were able to learn that the factors that contribute to the decision to stay or leave may be in play long before a conclusion is reached.

Acknowledgments

Preparation of this manuscript was supported by Grant 050106 from the National Institute of Mental Health (W. Furman, P.I.) and Grant 049080 from the National Institute of Child Health and Human Development (W. Furman, P.I.). Appreciation is expressed to the Project Star staff for their assistance in collecting the data, and to the Project Star participants and their partners, friends and families.

References

  1. Achenbach TM. Manual for the Young Adult Self-Report and Young Adult Behavior Checklist. Burlington, VT: University of Vermont Department of Psychiatry; 1997. [Google Scholar]
  2. Arnett JJ. Emerging adulthood: A theory of development from the late teens through the twenties. American Psychologist. 2000;55:469–480. http://dx.doi.org.du.idm.oclc.org/10.1037/0003-066X.55.5.469. [PubMed] [Google Scholar]
  3. Arnett JJ. Emerging adulthood: The winding road from the late teens through the twenties. New York: Oxford University Press; 2004. [Google Scholar]
  4. Beck AT, Rush AJ, Shaw BF, Emery D. Cognitive therapy of depression. New York: Guilford Press; 1979. [Google Scholar]
  5. Beyers W, Seiffge-Krenke I. Does intimacy precede identity? Testing Erikson’s theory on romantic development in emerging adults of the 21st century. Journal of Adolescent Research. 2010;25:387–415. doi: 10.1177/0743558410361370. [DOI] [Google Scholar]
  6. Bolger N, DeLongis A, Kessler RC, Wethington E. The contagion of stress across multiple roles. Journal of Marriage and the Family. 1989;51:175–183. doi: 10.2307/352378. [DOI] [Google Scholar]
  7. Collibee C, Furman W. Quality counts: Developmental shifts in associations between romantic relationship qualities and psychosocial adjustment. Child Development. 2015;86:1639–1652. doi: 10.111/cdev.12403. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Dailey RM, Rossetto KR, Pfiester A, Surra CA. A qualitative analysis of on-again/off-again romantic relationships: “It’s up and down, all around. Journal of Social and Personal Relationships. 2009;26:443–466. doi: 10.1177/0265407509351035. [DOI] [Google Scholar]
  9. DeMaris A. Till discord do us part: The role of physical and verbal conflict in union disruption. Journal of Marriage and the Family. 2000;62:683–692. doi: 10.1111/j.1741-3737.2000.00683.x. [DOI] [Google Scholar]
  10. Eggert LL, Herting JR, Thompson EA. The drug involvement scale for adolescents (DISA) Journal of Drug Education. 1996;26:101–130. doi: 10.2190/EQ6J-D4GH-K4YD-XRJB. [DOI] [PubMed] [Google Scholar]
  11. Erikson EH. Identity, youth, and crisis. New York: Norton; 1968. [Google Scholar]
  12. Felmlee D, Sprecher S, Bassin E. The dissolution of intimate relationships: A hazard model. Social Psychology Quarterly. 1990;53:13–30. doi: 10.2307/2786866. [DOI] [Google Scholar]
  13. Frost DM, Forrester C. Closeness discrepancies in romantic relationships: Implications for relational well-being, stability, and mental health. Personality and Social Psychology Bulletin. 2013;39:456–469. doi: 10.1177/0146167213476896. [DOI] [PubMed] [Google Scholar]
  14. Furman W, Buhrmester D. Methods and measures: The network of relationships inventory: Behavioral systems version. International Journal of Behavioral Development. 2009;33:470– 478. doi: 10.1177/0165025409342634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Furman W, Low S, Ho M. Romantic experience and psychosocial adjustment in middle adolescence. Journal of Clinical Child and Adolescent Psychology. 2009;38:1–16. doi: 10.1080/15374410802575347. [DOI] [PubMed] [Google Scholar]
  16. Gable SL, Gonzaga GC, Strachman A. Will you be there for me when things go right? Supportive responses to positive event disclosures. Journal of Personality and Social Psychology. 2006;91:904–917. doi: 10.1037/0022-3514.91.5.904. [DOI] [PubMed] [Google Scholar]
  17. Halford WK, Markman HJ, Kline GH, Stanley SM. Best practice in couple relationship education. Journal of Marital and Family Therapy. 2003;29:385–406. doi: 10.1111/j.1752-0606.2003.tb01214.x. [DOI] [PubMed] [Google Scholar]
  18. Halpern-Meekin S, Manning WD, Giordano PC, Longmore MA. Relationship churning in emerging adulthood: On/off relationships and sex with an ex. Journal of Adolescent Research. 2012;28:1660188. doi: 10.1177/074355841246524. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Harter S. Manual for the self-perception profile for adolescents. University of Denver; 1988. Unpublished manuscript. [Google Scholar]
  20. Harter S. Self-perception profile for adolescents: Revised manual and Questionnaires. University of Denver; 2012. Unpublished manuscript. [Google Scholar]
  21. Karney BR, Bradbury TN. The longitudinal course of marital quality and stability: A review of theory, methods, and research. Psychological Bulletin. 1995;118:3–34. doi: 10.1037/0033-2909.118.1.3. [DOI] [PubMed] [Google Scholar]
  22. Kessler RC, Berglund P, Demler O, Jin R, Merikangas KR, Walters EE. Lifetime prevalence and age-of-onset of DSM-IV disorders in the National Comorbidity Survey replication. Archives of General Psychiatry. 2005;62:593–602. doi: 10.1001/archpsyc.62.6.593. [DOI] [PubMed] [Google Scholar]
  23. Khaddouma A, Norona JC, Whitton SW. Individual, couple, and contextual factors associated with same-sex relationship instability. Couple and Family Psychology: Research and Practice. 2015;4:106–125. doi: 10.1037/cfp0000043. [DOI] [Google Scholar]
  24. Kogan SM, Lei M, Grange CR, Simons RL, Brody GH, Gibbons FX, Chen Y. The contribution of community and family contexts to African American young adults’ romantic relationship health: A prospective analysis. Journal of Youth and Adolescence. 2013;42:878–890. doi: 10.1007/s10964-013-9935-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Le B, Dove NL, Agnew CR, Korn MS, Mutso AA. Predicting nonmarital romantic relationship dissolution: A meta-analytic synthesis. Personal Relationships. 2010;17:377–390. doi: 10.1111/j.1475-6811.2010.01285. [DOI] [Google Scholar]
  26. Little TD. Longitudinal structural equation modeling. New York: Guilford Press; 2013. [Google Scholar]
  27. MacCallum RC, Zhang S, Preacher KJ, Rucker DD. On the practice of dichotomization of quantitative variables. Psychological Methods. 2002;7:19–40. doi: 10.1037/1082-989X.7.1.19. [DOI] [PubMed] [Google Scholar]
  28. Masten AS, Burt K, Coatsworth JD. Competence and psychopathology. In: Cicchetti D, Cohen D, editors. Risk, disorder and psychopathology (2nd ed). Developmental psychopathology. Vol. 3. New York: Wiley; 2006. pp. 696–738. [Google Scholar]
  29. Masten AS, Neemann J, Andenas S. Life events and adjustment in adolescents: The significance of event independence, desirability, and chronicity. Journal of Research on Adolescence. 1994;4:71–97. doi: 10.1207/s15327795jra0401_5. [DOI] [Google Scholar]
  30. Masten AS, Tellegen A. Resilience in developmental psychopathology: Contributions of the Project Competence Longitudinal Study. Development and Psychopathology. 2012;24:345–361. doi: 10.1017/S0954579412000003X. [DOI] [PubMed] [Google Scholar]
  31. Messer BJ, Harter S. Manual for the adult self-perception profile. University of Denver; 1986. [Google Scholar]
  32. Newcomb MD. Drug use and intimate relationships among women and men: Separating specific from general effects in prospective data using structural equation models. Journal of Consulting and Clinical Psychology. 1994;62:463–476. doi: 10.1037/0022-006X.62.3.463. [DOI] [PubMed] [Google Scholar]
  33. Neff LA, Karney BR. Stress and reactivity to daily relationship experiences: How stress hinders adaptive processes in marriages. Journal of Personality and Social Psychology. 2009;97:435–450. doi: 10.1037/a0015663. [DOI] [PubMed] [Google Scholar]
  34. Norona JC, Olmstead S, Welsh DP. Breaking up in emerging adulthood: A developmental perspective of relationship dissolution. Emerging Adulthood. 2016:1–12. doi: 10.1177/2167696816658585. [DOI] [PubMed] [Google Scholar]
  35. Ranta M, Dietrich J, Salmela-Aro K. Career and romantic relationship goals during emerging adulthood in times of economic uncertainty. Emerging Adulthood. 2014;2:17–26. doi: 10.1177/2167696813515852. [DOI] [Google Scholar]
  36. Raudenbush SW, Bryk AS, Congdon R. HLM 6 for Windows [Computer software] Skokie, IL: Scientific Software International, Inc; 2004. [Google Scholar]
  37. Raudenbush SW, Bryk AS. Hierarchical linear models: Applications and Data Analysis Methods. Thousand Oaks, CA: Sage Publishing; 2002. [Google Scholar]
  38. Rhoades GK, Kamp Dush CM, Atkins DC, Stanley SM, Markman HJ. Breaking up is hard to do: The impact of unmarried relationship dissolution on mental health and life satisfaction. Journal of Family Psychology. 2011;25:366–374. doi: 10.1037/a0023627. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Shulman S, Connolly J. The challenge of romantic relationships in emerging adulthood: Reconceptualization of the field. Emerging Adulthood. 2013;1:27–39. doi: 10.1177/2167696812467330. [DOI] [Google Scholar]
  40. Shulman S, Davila J, Shachar-Shapira L. Assessing romantic competence among older adolescents. Journal of Adolescence. 2011;34:397–406. doi: 10.1016/j.adolescence.2010.08.002. [DOI] [PubMed] [Google Scholar]
  41. Shulman S, Laursen B, Kalman Z, Karpovsky S. Adolescent intimacy revisited. Journal of Youth & Adolescence. 1997;26:597–617. doi: 10.1023/A:1024586006966. http://dx.doi.org.du.idm.oclc.org/10.1023/A:1024586006966. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Spielberger CD. State-Trait Anxiety Inventory (Form y) Redwood City, CA: Mind Garden; 1983. [Google Scholar]
  43. Sprecher S. The effects of self-disclosure given and received on affection for an intimate partner and stability of the relationship. Journal of Social and Personal Relationships. 1987;4:115–127. doi: 10.1177/0265407587042001. [DOI] [Google Scholar]
  44. Stanley SM, Markman HJ. Assessing commitment in personal relationships. Journal of Marriage and the Family. 1992;54:595–608. doi: 10.2307/353245. [DOI] [Google Scholar]
  45. Surra CA, Longstreth M. Similarity of outcomes, interdependence, and conflict in dating relationships. Journal of Personality and Social Psychology. 1990;59:501–516. doi: 10.1037/0022-3514.59.3.501. [DOI] [PubMed] [Google Scholar]
  46. Zimmer-Gembeck MJ, Hughes N, Kelly M, Connolly J. Intimacy, identity and status: Measuring goals in late adolescence and emerging adulthood. Motivation and Emotion. 2012;36:311–322. doi: 10.1007/s11031-011-9253-6. [DOI] [Google Scholar]

RESOURCES