Coping With Relationship Stressors: The Impact of Different Working Models of Attachment and Links to Adaptation

The study explores the role of working models of attachment in the process of coping with relationship stressors with a focus on long-term adaptation. In a 7-year longitudinal study of 112 participants, stress and coping were assessed during adolescence and emerging adulthood. In addition, working models of attachment were assessed by employing the Adult Attachment Interview (AAI). Individuals classified as having secure working models experienced low stress in relationships with parents, peers, and romantic partners and dealt with relationship stressors more actively by using their social network during adolescence and at the age of 21 years. In contrast, individuals with preoccupied working models experienced high relationship stress, particularly in relationships with parents, and employed less adaptive coping styles over time. Regression analyses revealed that although a preoccupied working model of attachment and withdrawal coping explained variance in symptomatology, relationship stressors were more predictive of poor psychological adaptation.

INTRODUCTION

In the last 2 decades, research on stress in adolescence has burgeoned (Grant et al., 2003). One important and consistent finding has been that normative, age-typical stressors are etiologically significant for understanding the trajectories of adolescent development that lead either to positive adaptation or result in psychopathology (Compas et al., 1995; Seiffge-Krenke, 2001). In this process, adolescent coping styles have been shown to be important as mediators and moderators of the impact of stress on current and future adjustment (Compas et al., 2001).

The coping styles that evolve during early and mid-adolescence are built on earlier experiences and influence how an individual will deal with new stressors occurring in late adolescence and emerging adulthood. Although there is abundant literature on coping with stress during adolescence (Compas et al., 2001; Sandler et al., 1997; Seiffge-Krenke, 1995), little work has specifically focused on relationship stress. This is surprising, since changes in most relationship domains are obvious and typical for this transition period. The developmental course of coping with relationship stress also remains unexplored. Little theory and research has been directed towards explaining this process and understanding the variables affecting its outcome in emerging adulthood.

A growing body of work suggests that attachment theory is a useful framework for understanding individual differences in coping with stress in adolescence (Howard and Medway, 2004; Ognibene and Collins, 1998) and its impact on future adaptation (Allen et al., 1998). Proponents of attachment theory (e.g., Bowlby, 1969, 1973) have highlighted the close links between a child's attachment to his or her primary caregiver and the child's competence in dealing with anxiety-arousing and conflicting social situations. It is theorized that, based on these early experiences, cognitive representations of self and others are developed. These representations, or working models of attachment, guide future coping behavior in times of stress, becoming relatively stable aspects of personality (Hesse, 1999).

Relationship Stress During Adolescence

Adolescence is a particularly challenging period in the life span. Most notably, adolescents experience rapid changes in their physical maturation. These changes coincide with the development of adolescent identity and the normative push for autonomy from parents. Adolescents seek to establish more mature and egalitarian relationships with their parents and must negotiate a balance between independence and dependence. As well, friendships become closer, an interest in entering romantic relationships increases, and social networks are enlarged to include romantic partners (Thornton et al., 1995). All of these factors may increase the pressure on the adolescent to revise his or her patterns of social interactions and hence, may result in conflicts in close relationships with significant others.

Indeed, research has shown that the vast majority of stressors experienced by adolescents (80%) pertain to relationships (Seiffge-Krenke, 1995). These result from everyday interactions that lead to conflicts within the family (Harvey and Byrd, 2000; Smetana et al., 1991; Seiffge-Krenke et al., 2001) or with peers and close friends (Anders and Tucker, 2000; Bowker et al., 2000) as well as issues associated with initiating and maintaining romantic relationships (Furman et al., 2002; Nieder and Seiffge-Krenke, 2001; Pollina and Snell, 1999).

According to Wagner and Compas (1990), whereas early adolescents report family-related stressors more often than other types of stressors, mid-adolescents name more stressors related to peers in their social network. Similarly, other studies have shown that mid-adolescents are greatly concerned by peer problems (Bowker et al., 2000; Seiffge-Krenke, 1995). Female adolescents name more conflicts with parents, peers, and romantic partners than males do (Phelp and Jarvis, 1994).

Relationship stressors and the ways the adolescent copes with them are closely intertwined. Difficulties in establishing autonomy and upholding relatedness with parents may forecast difficulties in competently interacting with peers (Furman et al., 2002) and in maintaining close, but individuated relationships with friends and romantic partners during adolescence and in emerging adulthood (O’Connor et al., 1996).

Coping Style

During the period of changing relationship patterns, adolescents’ attempts to manage stressful encounters with significant others are critically important. Similar to attachment behavior, which is activated in the context of perceived threat, stressful events usually trigger the use of coping strategies. Coping has been defined as an active, purposeful process of responding to stimuli appraised as taxing or exceeding the resources of the person (Lazarus, 1993) and includes behavioral, emotional, and cognitive attempts to manage the demands imposed by such stressors (Lazarus, 1998). Problem-focused or approach-oriented coping involves attempts to directly address the stressor, for example, by seeking support from others (Lazarus, 1993; Seiffge-Krenke, 1995). Cognitive-focused coping involves conscious reflection about the problem and how to resolve it (Garnefski et al., 2002; Seiffge-Krenke, 1995). Emotion-focused coping is characterized by attempts to regulate emotions or lessen emotional distress (e.g., by avoidance, minimization, distancing, or withdrawal) or at discharging it (self-blame or venting of emotion) (Lazarus, 1993; Seiffge-Krenke and Klessinger, 2000).

Lazarus (1998) has increasingly highlighted the importance of emotion-oriented coping strategies in a transactional context. The extent to which adolescents employ such strategies to regulate their behavior and emotions in stressful social interactions is critically important for relationship development. The regulation of negative emotions represents a central interface between attachment and coping, since dealing with negative affects such as fears, anxiety, or anger is learned through attachment experiences (Bowlby, 1973). Similarly, most stressful encounters with significant others demand emotion regulation in order to deescalate the situation and achieve self-maintenance (Kobak and Sceery, 1988). As such, coping is linked to but also distinct from attachment. While earlier attachment experiences may unconsciously guide stress perception and coping behavior, coping theory posits that coping is limited to those responses to stress that involve volition, effort, and conscious control (Compas et al., 2001; Lazarus and Folkman, 1984).

Coping responses vary systematically with how an individual appraises stressors. According to the Lazarus and Folkman (1984) model, stress appraisal involves the following 3 levels: (a) primary appraisal (e.g., evaluations of controllability and the magnitude of threat), (b) secondary appraisal (e.g., evaluations of the availability of own resources or social support for coping with the stressor), and (c) tertiary appraisal (e.g., evaluations of the effect of coping). Given that cognitive appraisal processes are likely to vary substantially with development, it seems warranted to suggest that coping behaviors will change over time as well. Unfortunately, little work has been done to determine the individual differences in stress appraisal during adolescence and how these differences are reflected in coping behaviors.

Working Models of Attachment

Attachment theory (Bowlby, 1969) and longitudinal research on attachment (Cassidy and Shaver, 1999) have helped to explain why individuals differ in the way they deal with stressors and to understand how these differences develop from childhood to adulthood. In stressful states, the attachment system is activated, and the degree of distress experienced by the individual varies as a function of his or her attachment style. Bowlby (1969, 1973) put forth the theory that through repeated experiences with attachment figures, children develop mental representations, or internal working models, which center on the availability and responsiveness of others, and the worthiness of the self. More specifically, these cognitive representations reflect (a) the availability and responsiveness of significant others (constituting a positive or negative model of others) and (b) the view of the self as competent (resulting in a positive or negative model of the self).

A central notion of attachment theory is that internal working models, once formed in early childhood, continue to function as relatively durable templates for later relationships throughout adolescence and into adulthood, where they help individuals to predict and manage stressful encounters, especially in relationships with significant others (George et al., 1985; Hesse, 1999). Such internal working models are critically important during adolescence, when individuals must appraise and deal with a considerable number of stressful encounters.

Attachment, Coping, and Adolescent Development

As detailed above, the conceptualization of working models of attachment put forward by Bowlby (1969, 1973) conjoin with Lazarus and Folkman's (1984) model of coping, according to which individual differences are believed to moderate the process of stress appraisal. Working models of attachment should be influential in the perception of stress (during the primary appraisal process) and, in particular, guide the individual during the secondary appraisal, when asking for and receiving support are crucial.

Research has indeed shown that adolescents with a secure working model perceive their families as more supportive (Harvey and Byrd, 2000; Kobak and Sceery, 1988) and demonstrate less anger, less avoidance, and more constructive engagement during conflicts with their mothers (Kobak et al., 1993). Similarly, securely attached young adults rely on their partners as sources of support and reassurance in an anxiety-provoking situation (Rholes et al., 1999) and are more likely to resolve conflicts by negotiating a solution and accepting compromises (Pistole, 1989). Individuals with a dismissing working model report little family support (Kobak and Sceery, 1988) and are less likely to seek social support outside the family (Ognibene and Collins, 1998). Individuals with a preoccupied working model of attachment use avoidance strategies to cope with interpersonal conflict more often, but are also more likely to seek support in response to conflict with mothers and dating partners (Howard and Medway, 2004; Torquati and Vazsonyi, 1999).

In summary, individual differences in stress perception and coping behavior due to attachment organization are to be expected. The question remains whether and how these differences are influenced by normative developmental changes. In this regard, it should be emphasized that from adolescence to emerging adulthood, the support-provider hierarchy changes in clearly identifiable ways (Furman et al., 2002). Furthermore, from early adolescence onwards, the onset of formal operations may permit greater diversity and flexibility in the range of coping strategies available to the individual (Kavsek and Seiffge-Krenke, 1996). Existing research seems to suggest that adolescents’ modes of coping with distressing social situations change with age. With increasing age, adolescents are more likely to utilize a greater variety and number of coping strategies (Bernzweig et al., 1993), take the perspective of significant others, and imagine the future consequences of an action (Garnefski et al., 2002). However, because of the cross-sectional nature of most research, developmental changes in coping style and cognitive appraisal processes are yet not fully understood, especially with respect to relationship stressors. These changes must be taken into account when analyzing the impact of attachment organization on stress perception and coping.

Links Between Stress Perception, Coping Style, Attachment, and Adaptation

In general, stress during the transition from adolescence to adulthood is a significant and pervasive risk factor for psychopathology (Compas et al., 1995). Although the accumulation of many stressors has been found to be directly related to psychopathology (Grant et al., 2003), the ways in which individuals cope with these stressors may have a bearing on current and future adjustment. The development of characteristic ways of coping with relationships stress may place adolescents on a more or less adaptive developmental trajectory. Several studies have shown that the use of maladaptive coping styles, such as avoidance, resulted in an increase of psychopathology concurrently and several years later (Seiffge-Krenke and Klessinger, 2000; Sandler et al., 1997; Seiffge-Krenke, 2001).

Individual differences in attachment organization may also influence the process of stress appraisal and subsequent coping. For example, a secure or insecure working model of attachment may determine which coping strategy will be used to deal with the stressor. Attachment organizations that lack coherence or security may create enduring vulnerabilities to psychopathology by impairing an individual's ability to cultivate and maintain close relationships, appropriately evaluate social interactions, and deal with stressors in these relationships competently. Insecure working models of attachment in adolescence have been linked to a variety of disorders, ranging from depression to problem behavior (Kobak et al., 1991; Allen et al., 1998). Similarly, insecure attachment organization in emerging adulthood has also been linked to severe psychopathology (Allen et al., 1996). As such, the moderating function of coping and attachment in buffering the effects of relationship stress on symptomatology needs to be explored.

This Study

Taken together, these findings seem to suggest that there are stable dispositions in coping behavior, depending on attachment organization. However, because of the cross-sectional nature of most research studies on this topic, it remains unclear (a) whether continuity in stress perception and coping behavior exists during adolescence, (b) whether continuity (or discontinuity) is related to a specific type of attachment organization, and (c) whether coping and attachment moderate the impact of relationship stress on future adaptation.

In the present 7-year prospective study, initiated when the participants were mid-adolescents and continued as they entered emerging adulthood, the links between stress perception, coping behaviors, working models of attachment, and adaptation were investigated, with a focus on relationship stressors. In accordance with developmental changes during adolescence, it was expected that whereas stress in relationship with parents would decrease, stress in relationships with friends and romantic partners would increase from mid- to late adolescence. It was further expected that adolescents would show changes in their coping styles in dealing with different relationship stressors. It was expected that active support-seeking would increase over time and that this increase would be particularly prominent when dealing with stress in friendships and romantic relationships. Further, it was hypothesized that reflection upon possible solutions, which incorporates the perspectives of significant others, would increase from adolescence to emerging adulthood, irrespective of the type of stressor at hand.

Other hypotheses were related to individual differences in stress perception and coping style, depending on attachment organization. On the basis of the assumption that individuals with a secure working model have a positive model of self and others, it was expected that they would report less stress in close relationships, compared to participants with preoccupied working models. In addition, it was expected that should relationship problems arise, participants with a secure working model would try to solve them by using active coping strategies (e.g., by talking with the person concerned and seeking support, if necessary), that they would reflect on possible solutions, thereby taking parents’, peers’, or romantic partners’ perspectives into account, and that they would show little avoidance behavior, compared to participants with insecure working models. In addition, it was hypothesized that the levels of active support-seeking and cognitive-oriented coping in participants with secure working models would increase over time, and that this increase would be more pronounced than for insecurely attached individuals. Taken together, it was expected that securely attached individuals’ preferred use of adaptive coping strategies to deal with relationship stressors would lead to an overall positive outcome, that is, less psychopathology, compared to individuals with dismissing or preoccupied working models of attachment.

In contrast, it was expected that individuals with dismissing working models would downplay relationship stressors and thus report significantly less stress in relationships with parents, peers, and romantic partners over time, compared to individuals with preoccupied working models of attachment. On the basis of the assumption that a dismissing working model of attachment incorporates a negative model of others, it was expected that dismissing individuals would be unlikely to seek support from others in times when relationship stressors arose; that is, their levels of active coping and support-seeking would be consistently low. Instead, it was expected that they would put great efforts into dealing with the stressor cognitively and show stronger gains in cognitive coping efforts over time, compared to individuals with other types of attachment models. Regarding overall adaptation, it was expected that by continuously showing a defensive downplay of stressors and being disinclined to seek support and address the person concerned, individuals with dismissing working models of attachment would be unable to reduce stress and thus show higher rates of psychopathology in the long run.

Finally, it was expected that individuals with preoccupied working models, because of the prevailing high levels of anger and conflict in their close relationships, would experience continuously high levels of stress in this domain across adolescence. It was expected that they would try to obtain comfort and help from parents, peers, and romantic partners but that they would be disappointed with the support offered to them. Therefore, it was expected that individuals with preoccupied working models of attachment would not make much progress in active coping and support-seeking over time, compared to those with secure working models. It was also expected that individuals with preoccupied working models would show consistently high scores in avoidance, compared to individuals with dismissing or secure models. Taken together, it was anticipated that they would exhibit equally high scores in active support-seeking and withdrawal. This highly ambivalent approach in dealing with relationship stressors would thus lead to an exacerbation of stress. Consequently, individuals with preoccupied working models of attachment were expected to have the highest risk for a maladaptive outcome.

METHOD

Participants

Sixty-four females and 48 males participated in this prospective longitudinal study. The mean ages of the participants at each of the 5 waves of the study were 14.05 years (SD=1.40), 15.03 years (SD=1.37), 16.02 years (SD=1.31), 17.02 years (SD=1.23), and 21.09 years (SD=1.39). These 112 participants belonged to a larger sample (N=145), 23% of which dropped out during the study and therefore did not participate in the AAI at the end. No significant differences emerged between participants who continued (n=112) and those who dropped out (n=33) with respect to sociodemographic characteristics such as age, gender, socioeconomic status (SES), educational background, parents’ marital status, or number of siblings. The final sample (n=112) was representative for German samples with respect to parents’ marital status (83% of the participants were from 2-parent families, and 17% lived with a single parent), number of children in the family (M=2.12, SD=0.77), SES (prevailing middle class), and nationality (93% of the sample were German nationals). The final level of schooling (53% of the participants received the Abitur, the highest level German high school diploma) and the proportion of those studying (46%) at age 21 compared to those in vocational training (54%) parallels the typical distribution for young Germans at this age.

Measures

The Adult Attachment Interview (AAI)

This structured interview was developed by George et al. (1985) to evaluate internal working models of attachment. The interview consists of 18 questions and requires about 1 h to complete. Participants were asked to describe their childhood relationships with their parents and to support their descriptions by providing particular memories. They were asked about instances of separation, rejection, threatening behavior, and being upset, hurt, or ill. In addition, they were asked to explain why their parents behaved the way they did, how these experiences had influenced their personalities, and what they had learned from the experiences.

In order to obtain measures of attachment, transcripts of the interviews were coded using the scoring system of Main and Goldwyn (1998). For this study, participants were assigned to 1 of 3 categories of working models, based on the individual's organization of thoughts, memories, and feelings, which is primarily manifested in verbal coherence criteria. The following 10 scales were used to evaluate attachment state of mind: (a) idealization of parent, (b) involvement of anger, (c) derogation, (d) insistence upon lack of recall, (e) metacognition, (f) passivity of discourse, (g) fear of loss, (h) unresolved loss, (i) overall coherence of transcript, and (j) overall coherence of mind. Values between 1 and 9 were defined for all scales, whereby a larger number always represented a high expression of the scale's described contents. In secure transcripts, participants were able to describe the relationships coherently, valued them, and found them influential in their lives. Dismissing transcripts were those in which the participants attempted to limit the influence of the relationship by idealizing, derogating, or failing to remember their experiences. In the preoccupied transcripts, the participants were vague, passive in speech, confused, angry, or preoccupied with the relationships.

The attachment interviews were coded according to the procedure described above by one coder, and a second coder reclassified 20 transcripts. Both coders had successfully completed Main and Hesse's Reliability Certification Procedure and were blind to all other data. The interrater agreement between the 2 raters was good (Kappa κ=0.81 across all scales). Reliability coefficients were as follows: idealization of parent, 0.67; involvement of anger, 0.74; derogation, 0.69; insistence upon lack of recall, 0.75; metacognition, 0.69; passivity of discourse, 0.71; fear of loss, 0.73; unresolved loss, 0.62; overall coherence of transcript, 0.79; and overall coherence of mind, 0.78.

Relationship Stressors

Relationship stress was measured by the Problem Questionnaire (PQ; Seiffge-Krenke, 1995), which assesses minor stressors in diverse domains. The instrument consists of 64 items that had been frequently named as typical and salient everyday stressors in earlier studies. The participants were asked to indicate the stressfulness of a specific problem, ranging from 1 (not stressful at all) to 5 (highly stressful). Factor analysis revealed the following 7 dimensions: academic stress, future-related stress, stress with parents, stress with peers, stress in leisure time, self-related stress, and stress in romantic relationships. Stressors belonging to the following domains were selected for this study: parents (10 items; sample item no. 19: “I fight with my parents because my opinions about many things differ from theirs”), peers (10 items; sample item no. 27: “I hardly have any friends”), and romantic relationships (7 items; sample item no. 49: “I am afraid that my jealousy could ruin my partnership”). Cronbach alphas for the 3 subscales were α=0.84, 0.83, and 0.74, respectively.

Coping With Relationship Stressors

The coping behavior of the participants during their adolescent years and as young adults was assessed with the Coping Across Situations Questionnaire (CASQ, Seiffge-Krenke, 1995). This instrument consists of a matrix of everyday stressors stemming from 8 domains (parents, peers, leisure time, romantic relationships, self, future, teachers, and school/studies/vocation) and 20 coping strategies. The participants were requested to mark all the coping strategies they used when a stressor in 1 of the 8 domains occurred. The coping strategies in each domain could be assigned to 1 of 3 coping scales: (a) active coping (7 items): coping strategies involving talking about the problem with the person concerned or seeking information, advice, or emotional assistance from formal (e.g., counseling centers) or informal support systems (e.g., parents and friends); (b) internal coping (6 items): coping strategies involving cognitive processes oriented towards searching for a solution, recognizing one's own limitations, and the willingness to accept compromises; and (c) withdrawal (7 items): efforts to withdraw from the stressor or avoid the problem as well as behaviors intended to reduce emotional tension (e.g., by screaming or slamming doors). In this study, active coping, internal coping, and withdrawal in the domains parents, peers, and romantic partners were used. The internal consistencies (alphas) for the 3 scales, summed across the domains, were 0.79, 0.82, and 0.80, respectively. To allow for direct comparison between the 3 different coping styles, mean values were established by dividing each scale by the number of items.

Symptomatology

The participants’ level of psychopathology was assessed in emerging adulthood by their responses to the Symptom Checklist 90—R (SCL-90-R, Derogatis, 1994). This instrument is a 90-item, self-report scale that measures multidimensional psychopathology, as rated according to a 5-point scale of distress (ranging from 0, not at all, to 4, extremely). In this study, the total score of symptomatology was used; Cronbach alpha amounted to 0.89.

Procedure

The participants were annually contacted at ages 14, 15, 16, and 17 years and requested to complete the PQ and the CASQ. An adaptation of the AAI as a measure for working models of attachment for adolescents was unavailable at the beginning of this longitudinal study; therefore, working models of attachment were assessed by conducting the AAI at age 21. In addition, the participants completed the CASQ and the SCL-90-R to assess symptomatology when they were 21 years old.

Table I. Means and Standard Deviations for Different Relationship Stressors in Individuals With Different Working Models of Attachment

Plan of Analyses

First, the distribution of working models as assessed by the AAI at age 21 was examined. Frequencies of secure, dismissing, and preoccupied groups were related to background and demographic variables. To examine the relationships between stress perception (assessed by the PQ) during adolescence (ages 14–17), coping behavior (assessed by the CASQ) during adolescence and young adulthood (ages 14–21), and internal working models (assessed by the AAI) at age 21, repeated measures MANOVAs were conducted, with the AAI classification and gender as between-subject factors. These first analyses showed that gender was not an important factor. Hence, in subsequent analyses, repeated measures MANOVAs were conducted with AAI classification as the between-subject factor and the 3 stressors (parents, peers, and romantic partners) and 3 coping styles used to deal with stressor involving parents, peers, and romantic partners (active coping, internal coping, and withdrawal), respectively, as dependent variables. In addition, repeated measures MANOVAs were conducted, with the total score of reported relationship stress, summed across the 3 stress domains, and the total score in each of the 3 coping dimensions, summed across the 3 relationship stressors. Differences between the attachment groups were tested by Bonferroni-corrected post hoc tests. Because of numerous comparisons, the alpha levels were adjusted. An ANOVA tested differences in symptomatology (based on the SCL-90-R) at age 21 between individuals with different working models of attachment. Finally, a stepwise multiple regression tested the variance explained in symptomatology at age 21 by the following predictors: (a) total score relationship stress, summed across time; (b) total score for active coping with relationship stress, summed across time; (c) total score for internal coping with relationship stress, summed across time; (d) total score for withdrawal coping with relationship stress, summed across time; and (e) the dummy coded scores of secure, dismissing, and preoccupied attachment classification. In addition, different interactions between total score relationship stress, coping with relationship stress, and attachment classification were included. A further regression was based on the combined score of coherence of representation and coherence of transcripts in the AAI.

RESULTS

Internal Working Models at Age 21

On the basis of the results of the coded interviews, 56 (50.0%) of the young adults were classified as having secure working models, 43 (38.4%) had dismissing working models, and 13 (11.6%) had preoccupied working models. The distribution of the male and female participants for the 3 working model groups followed the same pattern, χ²(2) = 0.58, ns. As well, the working models of individuals having different socioeconomic status (defined by the duration of the participant's education) did not differ, χ²(6) = 0.99, ns. However, compared to secure and dismissing individuals, more participants having a preoccupied working model came from single-parent families, χ²(2) = 10.80, p < 0.01.

Changes in Stress with Parents, Peers, and Romantic Partners in Individuals with Different Working Models

Table I provides descriptive statistics for the 3 relationship stressors over time for each of the 3 AAI groups. Regarding stress with parents, a significant main effect of attachment group emerged, F(2,110) = 4.900, p < 0.01, η² = 0.082, with preoccupied individuals reporting higher levels of stress with parents across all times (p = 0.009). Regarding stress with peers and romantic partners, significant main effects of time, F(3,330) = 3.149, p < 0.05, η² = 0.028 and F(3,330) = 4.988, p < 0.01, η² = 0.043, revealed a significant linear decrease of stress with peers and romantic partners in all individuals across time.

Changes in Coping With Stress With Parents, Peers, and Romantic Partners in Individuals With Different Working Models

Table II provides descriptive statistics for the 3 coping scales over time for each of the 3 AAI groups, broken down by relationship stressor. Regarding coping with stress with parents, the repeated measures MANOVA, with AAI classification as between-subject factors and the 3 coping dimensions as dependent variables, revealed a significant effect of time for active coping, F(4, 456) = 3.726, p < 0.01, η² = 0.032, with a cubic trend, F(1,114) = 7.830, p < 0.01, η² = 0.069. All individuals showed an increase in active coping with parental stress from ages 14 to 15, and, again, at age 21. A significant main effect of attachment group also emerged, F(2, 114) = 4.094, p < 0.05, η² = 0.067. Post hoc comparisons (Bonferroni) revealed a significant difference between the secure and the dismissing groups (p = 0.018), with a lower level of active coping in the latter group. Regarding internal coping with stress with parents, a Time × AAI group interaction emerged, F(6,342) = 2.257, p < 0.05, η² = 0.038, indicating that the increase in internal coping over time was much more prominent in the groups of participants with secure and dismissing working models, compared to the preoccupied working model group. With respect to the coping style withdrawal, a significant difference was found with respect to the AAI classification, F(2,114) = 3.011, p = 0.053, η² = 0.050, whereby participants with preoccupied working models tended to have higher scores than those with secure working models (p = 0.049).

Table II. Means and Standard Deviations in Active Coping, Internal Coping, and Withdrawal Coping With Relationship Stressors in Individuals With Different Working Models of Attachment

Table III. Intercorrelations Between the Study Variables

As for active coping with stress with peers, a Time × AAI interaction, F(8,452) = 2.188, p < 0.05, η² = 0.037, and a main effect of AAI group emerged, F(2,113) = 3.488, p < 0.05, η² = 0.058, with significant differences between participants in the secure and dismissing groups (p < 0.05). Whereas participants with secure working models showed a continuous increase in coping actively with stress with peers, participants with preoccupied working models exhibited stable high scores in active coping with peer stress. Regarding internal coping with respect to stress with peers, a Time × AAI group interaction, F(8,452) = 3.148, p < 0.01, η² = 0.052, and a marginally main effect of AAI group was found, F(2,114) = 3.037, p = 0.052, η² = 0.051. At age 14, the participants with a preoccupied working model showed higher scores in internal coping with peer stress, F(1,114) = 4.653, p = 0.011, than those with dismissing (p = 0.018) or secure models (p = 0.011). From ages 15 to 21, however, a main effect of AAI group emerged, F(2,114) = 5.731, p < 0.01, η² = 0.091, with participants in the secure working model group having higher scores in internal coping with peer stress than those in the groups with dismissing (p = 0.015) and preoccupied models (p = 0.037). Regarding withdrawal as a coping style, participants in the preoccupied working model group scored higher, F(2,114) = 3.989, p < 0.05, than those in the group with secure models (p = 0.017), but only at age 14.

Regarding coping with stress with romantic partners, a significant main effect of time with respect to active coping was found, F(4,404) = 9.709, p < 0.001, η² = 0.088, whereby the increase in this coping style was particularly prominent between ages of 14 and 15 in all groups. Similarly, all of the participants’ scores for internal coping with romantic stress increased over time, irrespective of the attachment group, F(4,404) = 3.760, p < 0.01, η² = 0.036. As for withdrawal, a Time × AAI group interaction emerged, F(8,404) = 2.167, p < 0.05, η² = 0.041, with a quadratic trend, F(2,101) = 5.301, p < 0.01, η² = 0.095. A main effect of AAI group was found, F(2,101) = 3.993; p < 0.05, η² = 0.073, with participants in the preoccupied group scoring higher in withdrawal than those in the secure (p = 0.021) and dismissing groups (p = 0.036).

Changes in Relationship Stress and Coping Over Time: The Impact of Working Models of Attachment

A MANOVA based on the total score of stressors in relationships revealed a significant time effect, F(3, 330) = 3.941, p < 0.01, η² = 0.035, with a cubic trend, F(1,110) = 5.975, p < 0.05, η² = 0.052. Over time, stress in relationships decreased, but not continuously. No difference with respect to attachment groups was found. A main effect of time emerged for active coping with relationship stress, F(4,404) = 5.862,  p < 0.001, η² = 0.055, illustrating a linear increase in this coping style in all individuals, F(1,101) = 16.798, p < 0.001, η² = 0.143. A significant main effect of attachment group was found, F(2,101) = 3.731, p = 0.027, η² = 0.069, and post hoc tests revealed that participants with dismissing working models scored significantly lower in this coping style than those with secure ones (p = 0.025). Regarding internal coping with relationship stressors, only a marginally significant effect of attachment classification was found, F(2,101) = 2.809, p = 0.065, η² = 0.053. Participants with secure working models tended to have higher scores in this coping style across time than those with preoccupied models (p = 0.072). Finally, a significant main effect of AAI group was found for the use of withdrawal to cope with relationship stressors, F(2,101) = 4.532, p = 0.013, η² = 0.082, whereby participants with preoccupied working models scored higher than those with secure models (p = 0.010).

Symptomatology, Depending on Working Models of Attachment

An ANOVA revealed significant differences between participants with different working models at age 21 F(2,104) = 4.949, p = 0.009. Participants with secure working models scored significantly lower (M = 42.48, SD = 28.05) than those with preoccupied models (M = 75.85, SD = 51.82, p = 0.034) but did not differ from those with dismissing models (M = 48.00, SD = 36.04).

Predicting Symptomatology by Stress, Coping, and Attachment Type

Intercorrelations between symptomatology at age 21 and the diverse predictors are displayed in Table III. Secure working models of attachment were negatively (r = −0.19) and preoccupied working models were positively (r = 0.29) associated with symptomatology. In addition, the correlations between active and internal coping and a secure working model were significant (r = 0.23 and r = 0.18, respectively), and preoccupied working models were associated with withdrawal coping (r = 0.28). Strong correlations existed between relationship stress and symptomatology (r = 0.56) and between withdrawal coping and symptomatology (r = 0.59).

Table IV. Results of Multiple Regression Analysis Predicting Symptomatology at Age 21 by Attachment, Relationship Stress Over Time, and Coping With Relationship Stress Over Time

Finally, a multiple regression analysis was conducted, with (a) gender and SES as control variables entered as a first step, (b) the dummy-coded scores of secure, dismissing, and preoccupied attachment classifications as a second step, (c) the total score for relationship stress, summed across time, as a third step, (d) the coping styles (i.e., the total score of active coping with relationship stress, summed across time, the total score internal coping with relationship stress, summed across time, and the total score withdrawal coping with relationship stress, summed across time) were entered and lastly the interaction terms. As can be seen in Table IV, 43% of variance was explained by this model. Neither gender nor SES predicted symptomatology at age 21. Relationship stress explained a substantial proportion (27%) and withdrawal 9% of the variance in symptomatology. Only the preoccupied working model of attachment was predictive of symptomatology (8% of variance explained). Neither interaction term including relationships stress, coping style, and attachment group contributed to higher variance explained than the main effect model. Another regression, based on the combined score of coherence of representation and coherence of transcript in the AAI, revealed similar results.

DISCUSSION

The first aim of the study was to analyze the developmental course of stress and coping during adolescence with a focus on relationships stress. In accordance with findings of Bowker et al. (2000) and Phelp and Jarvis (1994), all the participants in this study reported significantly higher stress in relationships with close friends and romantic partners, compared to stress with parents. Significant time effects further suggest that all adolescents, independent of attachment classification, experienced a decrease in stress with close friends and romantic partners across time. In contrast to the findings of Small et al. (1988) and Seiffge-Krenke (1999), who reported a decrease in stress with parents when children approach late adolescence, in the present study, comparably low stress levels with parents were retained across time.

How can we explain the initially high stress levels in close friendships and romantic relationships? Brown (1999) and Connolly and Goldberg (1999) described the developmental course of romance from adolescence to emerging adulthood in terms of phases that follow a specific sequence, whereby the first phase (initiation) seems to be particularly stressful. Also, both models acknowledge the influence that changes in the peer context may exert on the quality of romantic relationships over time. The dilemmas and anxieties related to group entry behavior have a bearing on how adolescents think about potential partners and what kinds of initiatives they take to approach them (Borja-Alvarez et al., 1991; Zarbatany et al., 1996). Nieder and Seiffge-Krenke (2001) studied the kinds of stress that young adolescents face in the beginning stages of romantic relationships. The relatively high levels of stress existing during this early phase of forming romantic relationships were found to be related to identity concerns, the peer context, and the romantic relationship itself. All adolescents in the present study perceived not having a boyfriend or girlfriend as highly stressful. Feeling insecure in dealing with the opposite gender and being afraid of losing contact with same-gender friends because of pairing up with a boyfriend or girlfriend were also considered as quite stressful. Since romantic attraction and encounters are initiated within the peer context (Connolly and Goldberg, 1999), these initiations may affect the relationships with friends. Together, these findings further our understanding of the stressors experienced during the initiation phase of romantic relationships and their ramifications on friendships.

The findings obtained in this longitudinal study further illustrate that stress related to interactions with peers and romantic partners decreases towards the end of adolescence, when romantic relations assume the characteristics of more enduring bonds (Seiffge-Krenke, 2003), and the functions of close friends and peers in supporting these relationships become more explicit (Brown, 1999). It was interesting to observe that over time, most participants in this study not only reported lower levels of peer-related and romantic stress, but coped with both relationship stressors more competently. Generally, from adolescence to emerging adulthood, active coping and support-seeking in dealing with relationship stressors significantly increased in all individuals, irrespective of problem at hand. Furthermore, the perspectives of others were increasingly taken into account when confronted with relationship stress, pointing to strong developmental changes in these 2 coping styles over the course of 7 years.

According to the premises of coping theory and attachment theory, working models of the self and others, which are developed as a result of early experiences with caregivers, should have a significant bearing on the appraisal processes in future stressful encounters (Bowlby, 1973; Lazarus and Folkman, 1984). The findings of the present study indeed suggest that attachment organization predisposes individuals to perceive and cope with relationship stress in particular ways. Working models of attachment were determined on the basis of the narratives of the AAI at 21 years. The distribution of internal working models in the sample (50% secure, 38% dismissing, and 12% preoccupied) corresponds to that of other nonclinical samples. In the present study, as in other investigations (see, for a review, Hesse, 1999), gender and socioeconomic status did not exert an influence on attachment representations. However, more participants with preoccupied working models of attachment came from single-parent families, compared to participants with secure and dismissing working models.

In this study, it became evident that the primary appraisal of stress depends on working models of attachment, especially with respect to stressors directly related to the attachment context, namely, stress with parents. Participants with preoccupied working models reported stable, high levels of stress with parents from ages 14 to 17. In contrast, although participants with secure and dismissing working models acknowledged that they had stress with their parents, their perceived levels of stress were considerably lower. Apparently, despite a normative amount of parent–adolescent conflict during adolescence, parents remain the primary secure base for these individuals (Harvey and Byrd, 2000).

This is further corroborated by the finding obtained in this study that participants with a secure or dismissing working model of attachment perceived stress with parents as being significantly less upsetting, compared to stress experienced with peers and romantic partners. In contrast, for the participants with a preoccupied working model, parents remained a major source of distress throughout adolescence. These participants showed stable high means for items, such as “I fight with my parents because my opinions about many things differ from theirs,” “I can't talk with my parents,” and “My parents don't have very much time for me.” Thus, participants with a preoccupied working model of attachment at age 21 had experienced continuous enmeshment, conflict, and actual anger with their parents as adolescents (from ages 14 to 17 years).

Coping with relationship stressors builds on earlier experiences. The characteristic ways of coping developed in childhood may be retained throughout adolescence and into adulthood and may position an individual on a more versus less adaptive developmental trajectory. In this study, participants with secure working models were expected to be better equipped to manage stressful situations with significant others. More specifically, their models of the self as being an active, competent agent, coupled with the perceptions that their interaction partners were supportive, would result in greater increases in active coping and support-seeking over time, compared to individuals with insecure working models.

The findings strongly support this hypothesis, since the participants with secure working models showed significant gains in their active coping and support-seeking when dealing with relationship stress from ages 14 to 21. As such, participants with secure working models were continuously more active in coping with stress with parents than were participants with dismissing working models, and they were continuously more active in coping with peer stress than participants with preoccupied working models. Their ability to cope competently was further evidenced by their increase in the use of internal coping over time, suggesting that they increasingly took the perspective of significant others into account. Moreover, participants with secure working models showed a consistently lower level of withdrawal when dealing with stress in close relationships with parents, friends, and romantic partners, compared to participants with preoccupied working models.

The findings obtained in this longitudinal study confirm and expand upon those obtained in cross-sectional studies which have shown that secure attachments are positively related to support-seeking and active problem-solving in adolescents (Greenberger and McLaughlin, 1998) and that young adults with secure working models experience higher levels of support from the family (Kobak and Sceery, 1988) and cope more actively with a variety of stressors, such as conflict and breakups in romantic relationships (Creasey and Hesson-McInnis, 2001; Davis et al., 2003). The findings of low symptomatology in participants with secure working models at age 21 further highlight a development which can be considered as beneficial in reducing future distress and leading to an overall positive outcome (Zeidner and Saklofske, 1996).

With respect to the impact of a dismissing working model of attachment on stress perception, coping style, and overall adaptation, the findings are less clear cut. Although theoretical and empirical work on attachment has suggested that individuals with dismissing working models of attachment tend to downplay stress (Bowlby, 1973; Cassidy and Shaver, 1999), the participants in the present study did not report less relationship stress over time than those with secure working models. However, participants with dismissing working models showed less active coping and support-seeking when dealing with relationship stress than participants with secure working models. Nevertheless, although they endeavored to deal with the stressor cognitively, their efforts did not exceed those of the participants with secure working models. Thus, the hypothesis that participants with dismissing working models of attachment would show a preference to use cognitive coping strategies in dealing with stress and that their use thereof would show a pronounced progression over time could not be fully confirmed.

Instead, the increases in the use of internal coping by participants with secure and with insecure working models of attachment are more likely to be explained by developmental factors. Increases in metacognitive skills, and, from mid-adolescence onwards, an improved ability to reflect on coping options and match coping efforts with perceived characteristics of the situation (Compas et al., 2001), as well as a greater capacity to reflect on the causes and consequences of conflicts in close relationships (Garnefski et al., 2002; Kavsek and Seiffge-Krenke, 1996) may account for these changes.

Finally, it was expected that participants with preoccupied working models would experience the highest stress levels and continue to try to obtain comfort and help in relationships when stressed, but that they would not make much progress in this coping style over time, due to disappointments with support given. As mentioned, participants with preoccupied working models continuously reported high stress levels with parents throughout adolescence. In addition, although their scores for active coping with diverse relationship stressors were quite high across time, from ages 14 to 21, their levels of withdrawal were continuously higher than those in the secure group. Thus, similar to the findings of Torquati and Vazsonyi (1999), participants with preoccupied working models in the present study showed an ambivalent coping style. They exhibited equally high scores in active support-seeking and avoiding the stressor across 7 years.

The findings discussed so far give rise to the question of whether working models of attachment should be conceptualized as a unified construct or not. Recent studies examining working models for parents (Furman and Simon, 2004), friends (Furman, 2001), and romantic partners (Furman et al., 2002) have shown that the different working models are distinct, yet related. In this regard, it is important to acknowledge that romantic relationships present new challenges and opportunities to the individual. This might be the reason why working models of parents and working models of peers were inconsistently related in the study conducted by Furman et al. (2002). The findings obtained in the present study support the idea of a distinct yet related association between diverse working models. Here the strongest impact of different working models was identifiable for stress with parents and coping with stress with parents. The decreasing amount of stress in friendships and romantic partnerships, together with the more active style in dealing with stress in these domains, suggest that individuals with different working models discriminate between types of relationship stressors and also adapt positively to changes in relationships.

Finally, the findings reported here support and elaborate upon the suggestion that insecure working models of attachment may result in the use of maladaptive coping strategies in response to stress (Allen et al., 1996; Kobak et al., 1991). At first sight, the focus on cognitive coping exhibited by participants with a dismissing working model can be considered as adaptive. However, as the Kraaij et al. (2003) study demonstrated, a coping style which is more or less limited to the use of cognitive strategies may be considered as risk factor for depressive symptoms. Furthermore, not seeking support in stressful times is considered to be a risk factor for health (Kotler et al., 1994). Although participants with a dismissing working model did not show elevated levels of symptomatology at age 21, in the long run their focus on cognitive coping might turn out to be a risk factor.

The finding obtained in this longitudinal study that participants with preoccupied working models showed continuous high levels in withdrawal from adolescence to young adulthood confirms and expands upon findings from cross-sectional studies in which insecurely attached adolescents used more avoidant coping (Howard and Medway, 2004; Torquati and Vazsonyi, 1999). Earlier research has stressed the long-term consequences of avoidant coping for a problematic outcome. Individuals who rely on this coping strategy show a prognosis of increased psychopathology that may last from 1 year (Herman-Stahl et al., 1995) to several years (Seiffge-Krenke and Klessinger, 2000).

Limitations

One limitation of this study was that internal working models were assessed only once. Although adult attachment styles are presumed to reflect relatively stable and enduring orientations, they are potentially modifiable by discomforting life events and experiences (Bowlby, 1973; Hesse, 1999). Moreover, from an attachment perspective, adolescence is a transitional period (Allen and Land, 1999). Future longitudinal studies assessing internal working models at different time points during adolescence and adulthood could establish the stability of internal working models of attachment.

Other limitations of this study primarily involve the use of self-reports and the sample size. First, data were obtained from the same source and by means of the same method (self-report questionnaires and interviews). Using a second source and incorporating observational methods, for example, in order to assess different coping styles in a relationship context, would be advisable. Second, owing to the small samples in the 2 groups of participants with insecure working models of attachment, the findings have to be interpreted with caution. The findings obtained in this study showed that dismissing and preoccupied working models exert a differential influence on coping with relationship stress and that their coping style also develop differently over time. Although the findings overall were in accordance with most hypotheses, some of the numerous comparisons were marginally significant. This necessitates replications using larger samples, in particular, with a greater number of participants with a preoccupied working model. Finally, future analyses of the participants’ symptomatology during their adolescent years might provide more information that would answer the question of whether the links between stress perception, coping style, and symptomatology shown by the participants with preoccupied working models might have existed already early on.

CONCLUSIONS

Understanding the dynamics of stress and coping in the lives of adolescents is of both theoretical and practical significance. This study focused on frequent and changing stressors during adolescence involving relationships with parents, peers, and romantic partners and explored the role of 2 variables, attachment and coping, both of which are activated in the context of perceived threat. The development of abilities to cope in changing relationship contexts was acknowledged and related to dispositions rooted in attachment experiences.

The findings highlight that there are developmental changes in stress perception and coping behavior, and emphasize how relationships with significant others change at the transition to adulthood. The findings documented that whereas the vast majority of participants coped adaptively with relationship stress from adolescence to emerging adulthood, individual differences placed some participants on less adaptive developmental trajectories. A preoccupied working model of attachment can be considered as a risk factor for an unfavorable outcome. This type of working model was linked with stable, maladaptive coping styles from adolescence onwards and led to increased symptomatology in emerging adulthood. This finding is a good starting point for intervention. This study further shows that, with respect to prevention, individuals with a dismissing working model of attachment can be considered at risk for psychopathology in the long run.