Journal of Abnormal Child Psychology

, Volume 39, Issue 3, pp 389–400

Emotional Dysregulation and Interpersonal Difficulties as Risk Factors for Nonsuicidal Self-Injury in Adolescent Girls


    • Department of EpidemiologyUniversity of Washington, Child Health Institute
  • Janice Zeman
    • Department of PsychologyCollege of William and Mary
  • Cynthia Erdley
    • Department of PsychologyUniversity of Maine
  • Ludmila Lisa
    • Department of Psychiatry and PsychologyMayo Clinic
  • Leslie Sim
    • Department of Psychiatry and PsychologyMayo Clinic

DOI: 10.1007/s10802-010-9465-3

Cite this article as:
Adrian, M., Zeman, J., Erdley, C. et al. J Abnorm Child Psychol (2011) 39: 389. doi:10.1007/s10802-010-9465-3


The purpose of this study was to examine a model of factors that place psychiatrically hospitalized girls at risk for non-suicidal self-injury (NSSI). The role of familial and peer interpersonal difficulties, as well as emotional dysregulation, were examined in relationship to NSSI behaviors. Participants were 99 adolescent girls (83.2% Caucasian; M age = 16.08) admitted to a psychiatric hospital. Structural equation modeling indicated the primacy of emotional dysregulation as an underlying process placing adolescents at risk for NSSI and mediating the influence of interpersonal problems through the family and peer domains. When family and peer relationships were characterized by conflict and lack of support for managing emotions, adolescents reported more dysregulated emotion processes. Family relational problems were directly and indirectly related to NSSI through emotional dysregulation. The indirect processes of peer relational problems, through emotional dysregulation, were significantly associated with NSSI frequency and severity. The findings suggest that the process by which interpersonal difficulties contribute to NSSI is complex, and is at least partially dependent on the nature of the interpersonal problems and emotion processes.


Emotion regulationNonsuicidal self-injuryPeer relationsParent-child relations

The transition to adolescence is associated with substantial increases in a wide range of problematic behavior including delinquency, drug use, anxiety, depression, and suicidal behaviors (Cicchetti and Toth 1998; Loeber and Farrington 2000). Nonsuicidal self-injury (NSSI), the purposeful, direct destruction of body tissue without conscious suicidal intent (Suyemoto 1998), is a troubling behavior that increases in prevalence during adolescence (Ross and Heath 2002). Although there is general consensus that NSSI begins in early adolescence, less research has focused on NSSI in adolescents as compared to adults. Given that NSSI has emerged as a formidable threat to physical health and psychological functioning in adolescence (O’Loughlin and Sherwood 2005), research is needed to clarify the etiology of NSSI. Linehan’s (1993) biosocial model posits that the etiological mechanisms that contribute to the development of emotional dysregulation and borderline personality disorder involve the interaction of biologically-based vulnerability to intense emotionality with an inadequate environment for learning to manage emotion. In a recent extension of Linehan’s (1993) original biosocial model, Crowell et al. (2009) suggest that emotional dysregulation fosters and maintains self-injury within an adversarial and unsupportive social context. Thus, the present study sought to test this hypothesis by examining the associations between the occurrence of NSSI and two developmentally important social contexts of adolescence: parental and peer relations.

The study focused on a clinical sample of girls due to research indicating that adolescent girls are particularly at risk for psychological distress during adolescence (Hilt et al. 2008), and NSSI predominately occurs within the context of psychological distress (Nock and Prinstein 2005). Although NSSI does occur in boys, studies suggest a predominance of this behavior in adolescent girls. For example, Nixon et al. (2002) reported that 86% of adolescents in inpatient and partial hospitalization settings who participated in NSSI were girls. Further, a prospective follow-up study in a community sample (Sourander et al. 2006) found that at age 12, 2.7% of girls and 3.1% of boys reported engaging in NSSI. By age 15, 12.6% of girls and 4.1% of boys reported involvement in NSSI. Thus, NSSI is relatively rare before age 15, but increases through adolescence, particularly for girls.

Emotional Regulation Function of NSSI

A number of functions of NSSI have been proposed, with the emotion regulatory function of this behavior receiving the most empirical support (see Klonsky 2007, for a review). Emotion regulation (ER) abilities are characterized by the processes that are employed to modify emotional experiences (Thompson 1994). Research suggests that the construct of ER is a useful heuristic for defining and understanding risk factors for the development, maintenance, and exacerbation of childhood psychopathology (Cicchetti et al. 1995). Individuals who are emotionally dysregulated exhibit patterns of responding in which there is a mismatch between their goals, responses, and/or modes of expression, and the demands of the socio-cultural context (Zeman et al. 2006). Individuals may manage emotional intensity with behaviors, such as NSSI, that serve the desired function of decreased emotion, but have detrimental consequences (Gratz 2003).

The functionalist theory of emotion posits that the processes involved in emotional responding are context-dependent and shaped by an individual’s goals within social contexts and experiences within close relationships (Saarni et al. 1998). In general, features of emotional predictability, openness to emotion, and the role of adults helping to restore emotional homeostasis appear to assist adolescents in positive emotional development (Larson and Brown 2007). In contrast, the development of emotion dysregulation is posited to occur through the interaction between the vulnerability to high-intensity emotion and an inadequate learning history for managing emotional intensity in constructive ways (Bradley 2000; Crowell et al. 2009). Interpersonal difficulties in significant relationships may be an important variable that interferes with learning new or modifying former ineffective ER strategies (Saarni et al. 1998). Consistent with this conceptualization, the present study characterized parental relationship problems as (a) punishment, minimization, or denial of the adolescent’s negative emotion, (b) high levels of parent-adolescent conflict, and (c) low levels of socio-emotional support. Peer relational problems were conceptualized as processes that disregard an adolescent’s well-being through (a) victimization within the peer group, and (b) low quality friendship. In sum, relationships characterized by conflict and lack of support may serve to de-legitimize valid emotional experiences, and interfere with learning opportunities to regulate affect adaptively in accord with the demands of the social context.

Adolescent socio-emotional development occurs primarily within family and peer contexts as youth are provided with learning opportunities through repeated interactions within these relationships. Parental figures are the first socializing agents and shape the emotional trajectory through direct and indirect methods. Not surprisingly, individuals raised in harsh and conflictual family climates have an array of emotional competence deficits (Fruzzetti et al. 2005). Environments characterized by conflict and aggression and relationships that are cold, unsupportive and/or neglectful represent vulnerabilities that are suspected to disrupt emotional processes and psychological functioning (Repetti et al. 2002). Poor quality family relationships have been examined broadly in relation to suicidal behaviors (e.g., Johnson et al. 2002). However, most research investigating the risk factors for NSSI has focused on retrospective accounts of early childhood experiences within the family. As an exception, a recent examination extended these retrospective results and provided more compelling support for disrupted family interaction (Crowell et al. 2008). Specifically, in a conflict discussion task, families of self-injuring adolescents demonstrated less positive affect, more negative affect, and lower cohesiveness compared to non-injuring controls. Thus, theory and emerging empirical findings support, at minimum, family dysfunction as concomitant with NSSI.

As children enter early adolescence, peer settings become increasingly salient and provide potentially powerful opportunities for emotional development. Although peer relations are often thought to positively impact emotional development (e.g., validation, support), some children have aversive experiences with peers that can contribute negatively to emotional development. Victimization and poor friendship quality are two such forms of difficulties that have been shown to interfere with adolescents’ emotional functioning. Peer victimization is characterized by repeated negative confrontations between a perpetrator and targeted peer (Olweus 1978). This broad definition can include physical (e.g., hitting), verbal (e.g., teasing), and indirect (e.g., exclusion from a group) forms of harassment. Victimization has been linked with adverse social and emotional outcomes (e.g., loneliness, depression, social anxiety) for children and adolescents (see Hawker and Boulton 2000, for a review), and its connection to NSSI in adolescents has been investigated in only one study to date. Hilt et al.’s (2008) examination of victimization in Hispanic and African American young adolescent girls revealed that peer victimization was specifically related to social functions of NSSI (e.g., using NSSI to obtain a response from another person). In contrast, internalizing symptoms were associated with adolescent girls endorsing emotion regulatory functions of NSSI (e.g., reducing negative emotions). Providing further evidence for the important role of peer relations, the association between peer victimization and NSSI was moderated by peer communication. Those girls who endorsed low quality peer communication were likely to engage in NSSI when reporting higher levels of peer victimization. This study suggests that specific qualities of peer interaction play a role in NSSI and warrant further investigation.

Another significant aspect of peer experience, particularly during adolescence, is friendship, defined as a close, dyadic mutual relationship with a shared history (Bukowski and Hoza 1989). Adolescents who lack a close, supportive friendship are at risk for difficulties in psychosocial adaptation, including problems coping with severe life stressors (Cohen and Wills 1985) and suicidal behaviors (Spirito et al. 1989). Notably, among children who do have friends, the quality of their friendship experience (e.g., level of intimacy, validation) varies (Parker and Asher 1993). Children with low quality friendships are more likely to experience loneliness and depression (Nangle et al. 2003). Whereas several studies have revealed links between friendship experiences and certain aspects of psychological adjustment, there is little research on how low quality friendships may relate to NSSI. There is, however, research that has illustrated the importance of friendship in suicidal behaviors. For example, for children under the age of 16, almost 40% indicated that friendship difficulties were the main precipitant to their suicide attempts (Hawton et al. 1996). Further, social support from friends was found to moderate the association between depression and later suicidal behavior, such that teens in high school who were very depressed but perceived a high level of friendship support were less likely to exhibit later suicidal behavior than those who were depressed and endorsed a low level of friendship support (Reifman and Windle 1995). Given the developmental significance of friendship in adolescence, its well-established connection to social-emotional functioning, and findings that friendship quality is related to risk for suicidality, it seems likely that friendships that are characterized by low levels of support and high levels of conflict may increase adolescents’ vulnerability to NSSI.

The Current Study

The primary purpose of this study was to examine a model of factors that place adolescent girls at risk for engaging in NSSI. Specifically, the role of ER, family difficulties (i.e., parental punishment, minimization, or denial of child’s negative emotion; high levels of conflict; low levels of support), and peer difficulties (i.e., peer victimization and friendships characterized by negative interactions) were explored. Both theory and preliminary research findings suggest that there may be a meaningful relationship among emotion dysregulation, interpersonal problems, and NSSI (Crowell et al. 2009; Sim et al. 2009). Briefly summarized, the experience of living in an environment characterized by an enduring pattern of parental relationship difficulties may limit an adolescent’s capacity to cope adaptively with intense emotional experiences. During times of stress, individuals raised in such environments are likely to have difficulty accurately identifying and expressing their emotions, which then places them at risk for reducing strong emotion by relying on avoidant coping strategies, such as NSSI. Emotional patterns developed through the parental relationship may confer risk in the peer world. Specifically, the emotional exchange that occurs between parents and their children has direct effects on the emotional aspects of the interactional style that children have with their peers (Carson and Parke 1996; Eisenberg et al. 2001). Difficulties in peer contexts may compound prior relational difficulties and contribute to problems with ER and psychological adjustment. Although much less research has investigated peer socialization of emotion, it is likely that peers are a powerful agent in this realm (Perry-Parrish & Zeman in press). Thus, this study examined both parental and peer processes that may contribute to NSSI.

To better understand the mechanisms of NSSI, this study tested the hypothesis that the risk for NSSI increases when the individual experiences emotional dysregulation within an unsupportive social context. Thus, a model of several risk factors (i.e., emotional dysregulation, interpersonal difficulties in the family and peer contexts) that appear to contribute to increased vulnerability for engaging in NSSI was evaluated. Emotion dysregulation has been implicated in diverse forms of adolescent psychopathology, including both internalizing and externalizing problems (see Zeman et al. 2006). Given that NSSI is prevalent in both internalizing and externalizing problems (Nock et al. 2006), emotional dysregulation was conceptualized as an underlying feature of disordered behavior and thus, reliance on and associations with diagnostic classification was not a goal of this research. Data were collected via self-report questionnaires and the use of an adolescent emotional reflection task because the merits of assessing internal processes through self-report outweigh the possible limitations, particularly in adolescence (Inderbitzen 1994).

Based on theory and empirical findings, it was predicted that (a) emotion dysregulation would be positively correlated with engagement in multiple methods and frequency of NSSI, (b) environments characterized by interpersonal difficulties (parental punishment, minimization, or denial of child’s negative emotion; poor friendship quality; peer victimization) would be positively associated with emotional dysregulation, and (c) social environments characterized by interpersonal difficulties would be positively associated with NSSI frequency. It was hypothesized that the association between interpersonal difficulties and NSSI would be partially mediated by emotional dysregulation. Additionally, it was expected that there would be direct effects between interpersonal difficulties and NSSI. Given the different functions of the parent and peer social environments, it was hypothesized that interpersonal difficulties experienced within each domain would predict uniquely to dysregulation. Based on the literature that suggests linkages between family and peer functioning and that family socialization, particularly ER, precedes and forms the basis for later peer socialization, it was also expected that interpersonal difficulties within the family would predict problematic peer relationships.



Participants included 99 adolescent girls (ages 13.5–18.5 years, M = 16.08; SD = 1.42) who were consecutively admitted to a child and adolescent psychiatry unit over a 12-month period and who were part of a larger study on emotional functioning and psychological adjustment in adolescents (Adrian et al. 2009). The participants were predominantly Caucasian (83.2%), followed by Hispanic (13%), Asian (2.3%), and African American (1.5%). The sample was primarily composed of middle-class social status (Hollingshead 1975; M = 2.21, index of social position=44.99). To characterize the sample, diagnoses are provided with exclusion criteria including psychotic symptoms and/or developmental disabilities. Upon admission, a psychiatry resident formulated a preliminary diagnosis through a brief psychosocial assessment for the purposes of acute crisis stabilization and treatment planning. This process indicated that 90% of patients were admitted for suicidality and accordingly, many automatically received a diagnosis of unipolar depression. Through patient and family interview, observation, and interdisciplinary meetings, diagnoses were re-evaluated and finalized at discharge. Regarding the primary discharge diagnosis, the majority of the sample had a mood disorder (72.1%) followed by an adjustment disorder (6.4%), an anxiety disorder (6.4%), an eating disorder (5%), a disruptive behavior disorder (2.9%), ADHD (2.1%), substance abuse (1.4%), or a somatoform disorder (0.7%), with 2.9% not diagnosed. Some adolescents (40%) had a secondary diagnosis. Through an extensive chart review of the social work report and dismissal summary, we discovered that 92% of the sample had at least one notable externalizing symptom, 87% had at least four externalizing behaviors, 37% had significant drug or alcohol problems, and 67% had parent-child conflict or family dysfunction.


Emotion Dysregulation

The Difficulties with Emotion Regulation Scale (DERS; Gratz and Roemer 2004) is a 36-item self-report measure assessing individuals’ characteristic patterns of emotion dysregulation. The DERS consists of six subscales (Nonacceptance of Emotional Responses, Difficulties Engaging in Goal-Directed Behavior, Impulse Control Difficulties, Lack of Emotional Awareness, Limited Access to Emotion Regulation Strategies, and Lack of Emotional Clarity) that were theoretically formulated and confirmed through factor analysis. Participants rated responses on a 5-point Likert scale (1=almost never, 5=almost always). Higher scores indicate greater emotional dysregulation. Internal consistency of the total scale was strong in this study (α = 0.93) with each subscale greater than 0.80, consistent with previous research (Gratz and Roemer 2004).

The Emotion Expression Scale for Children (EESC; Penza-Clyve and Zeman 2002) is a 16-item questionnaire that assesses two aspects of deficient emotional expression. The Poor Awareness scale measures lack of emotional awareness, whereas the Expressive Reluctance scale examines the lack of motivation to express emotion to others. Items are rated on a 5-point Likert scale (1=not at all true, 5=extremely true). Validity has been documented by the scales’ positive correlations with sadness and anger inhibition and dysregulated expression, and negative correlations with ER coping (Zeman et al. 2002). In the present study, the reliability was adequate (Poor Awareness = 0.85; Expressive Reluctance = 0.83).

Family Interpersonal Problems

Family relational problems were conceptualized as parental responses to the child that were characterized by the non-acceptance of emotions, high levels of conflict, and few opportunities for support and closeness with the caregiver. This conceptualization is similar to theoretical accounts (e.g., Crowell et al. 2009) and empirical measurement (e.g., Selby et al. 2008) of family relational difficulties. For this study, family interpersonal difficulties were assessed using the Emotions as a Child questionnaire and the Family Environment Scale.

The Emotions as a Child (EAC; Magai 1996) is a 45-item questionnaire that measures children’s perceptions of their emotion management strategies and their perceptions of how their primary caregiver responds to their emotions. Participants were asked to rate, on a Likert-type scale (1=not at all typical, 5=very typical), their responses to sadness, anger, and fear as well as their perception of their caregiver's behavior in response to these emotions over the past year. The scales were condensed into two broad emotional socialization strategies: supportive (reward; α = 0.71) and unsupportive (punish, neglect, override; α = 0.62) across the three measured emotions.

The Family Environment Scale (FES; Moos and Moos 1994) was completed by adolescents to evaluate their family’s social and emotional environmental characteristics. Although the FES has three dimensions, this study used only the Cohesion and Conflict subscales. Participants responded to the 18 items on a dichotomous true or false scale. The FES has adequate psychometric properties including construct, content, and predictive validity with normative samples and with those families who experience depression and alcoholism (Moos and Moos 1994). In this study, reliability was acceptable (Cohesion = 0.72; Conflict = 0.77).

Peer Interpersonal Problems

Peer relational problems were conceptualized as peer responses that are characterized by non-acceptance of adolescents’ characteristics and lack of support or concern for their emotional well-being. This construct was assessed by two well-validated constructs in the peer relations literature, victimization and friendship quality.

The Social Experiences Questionnaire- Self-Report version (SEQ-S; Crick and Grotpeter 1996) was used to assess participants’ experiences of both relational and overt victimization. The SEQ-S consists of 12 items that load onto three subscales. The Relational Victimization scale assesses the frequency with which peers damage friendships, acceptance, and peer group inclusion. The Overt Victimization scale measures the frequency with which peers threaten or harm a person’s physical well-being. The Recipient of Prosocial Acts scale assesses the frequency with which the participant is a target of caring acts from peers. Adolescents responded to each question on a 5-point Likert-type scale (1=never, 5=all the time). The SEQ-S has demonstrated adequate internal reliability with alphas from 0.84 to 0.88 in this study.

The Network of Relationship Inventory (NRI; Furman and Buhrmester 1985) was used to assess the features of the participant’s best friendship. The measure consists of 10 friendship dimensions including companionship, conflict, instrumental aid, antagonism, intimacy, nurturance, affection, admiration, relative power, and reliable alliance. Responses to the 30 items are rated on a 5-point Likert scale (1=little or none, 5=the most) and load onto two factors. The Social Support factor is the average of the companionship, instrumental aid, intimacy, nurturance, affection, admiration, and reliable alliance scores. The Negative Interactions factor is the average of the conflict and antagonism items. Internal consistency was adequate in this study (negative interactions = 0.89, friendship support=0.86).

Nonsuicidal Self-Injury and Psychological Functioning

To understand the frequency and severity of self-harm behaviors, adolescents completed an adapted version of the self-report measure, the Self-Harm Behavior Questionnaire (SHBQ; Gutierrez et al. 2001). The SHBQ is comprised of four parts, but for this study only the first section was used. This section asks about intentional self-injury that was not suicidal in nature and obtains information about the type of behavior, frequency, intent, lethality, and outcome. Internal consistency was adequate with an alpha of 0.91 for the self-harm section.

The Youth Self Report is a parallel version of the CBCL (Achenbach and Rescorla 2001) for adolescents aged 11–18. The YSR has 112 items rated on a 3-point scale and yields eight syndrome scales and three broad-band categories relating to internalizing, externalizing, and total problems. Internal consistency of the CBCL subscales was adequate (0.81 for internalizing problems; 0.87 for externalizing problems).


After obtaining parental consent and adolescent assent, in a 45-minute session, girls completed the measures on a computer in a private room with a research assistant available to answer questions or assist with the technology. Then, girls completed the emotional reflection task to assess their emotional awareness and emotional language skills. Participants were told: “We would like you to write about a recent experience where you felt strong emotion. There are no right or wrong answers. We just want you to write about what happened, how you felt, and how you coped with the situation and those feelings. You will have about 2 min. You can choose to write on paper or use the computer.” The reflection tasks were scored using Linguistic Inquiry and Word Count (LIWC; Pennebaker et al. 2007a), which is a text analysis software that calculates dimensions of language use including emotion words, social words, and self-referencing. For the purposes of this study, the primary measure was number of emotion words utilized in the writing sample. The LIWC evidences strong reliability (positive words, adjusted α = 0.97; raw = 0.40; negative emotion words, adjusted α = 0.97, raw = 0.61; Pennebaker et al. 2007b). Regarding the validity of the LIWC coding system, those who use high levels of positive and negative emotion words are more likely to experience positive outcomes, whereas those who do not use emotion words when describing traumatic events are at greatest risk for subsequent psychological and physical health problems (Pennebaker et al. 2003).


Preliminary Analyses

Assumptions of normality were first evaluated. The results indicated that some variables were right skewed (i.e., NSSI frequency, NSSI methods, number of emotion words in writing sample, overt and relational victimization, negative friendship interactions) and subsequently were treated with square root (i.e., NSSI frequency, NSSI methods, number of emotion words correction words) or logarithmic (i.e., overt and relational victimization, negative friendship interactions) transformations. These transformations effectively normalized the data and were used in subsequent analyses. As can be seen in the correlation matrix (Table 2), many of the measures are correlated. However, the correlations were not so high that they created problems with inverting the covariance matrix, and the determinant of the covariance matrix did not indicate a problem with multicollinearity.

Descriptive Analyses

Seventy-seven (78.5%) of the participants reported engagement in NSSI (see Table 1). Girls who said they had engaged in NSSI endorsed the following behaviors: cutting skin (74.7%), rubbing skin (36.4%), hitting self (34.3%), stabbing/puncturing skin (27.3%), and burning skin (22.2%). The mean number of NSSI behaviors reported by adolescents was 2.62 (SD = 1.35, range = 1–5). The preferred method of NSSI was cutting; however, the majority of participants engaged in more than one NSSI behavior. Although the overall level of lethality was low, the entire range of lethality from no intervention to inpatient medical care was documented. Compared with adolescents who did not endorse engaging in NSSI, adolescents who did engage in NSSI had significantly more emergency room visits, suicide attempts, and greater severity on self-reported psychopathology (internalizing, externalizing, and total problems on the YSR), and they were more likely to be diagnosed with a mood disorder. See Table 2 for correlational matrix and descriptive statistics.
Table 1

Demographic information by self-injury status


Non-self-injuring adolescents (N= 22)

Self-injuring adolescents (N= 77)

Significance test

Effect size




t (98) = 1.99

Cohen’s d = 0.26




t (96) = −1.31

Cohen’s d = 0.25

Achenbach YSR Scale

Internalizing Problems


66.68 (8.49)

t (96) = −3.65***

Cohen’s d = 0.64

Externalizing Problems

56.69 (13.04)


t (96) = −3.14***

Cohen’s d = 0.56

Total Problems

53.31 (9.80)

67.34 (7.91)

t (96) = −4.55***

Cohen’s d = 0.85

Global Assessment of Functioning

35.03 (6.16)

34.94 (9.30)

t (96) = 0.05

Cohen’s d = 0.01

Mood Disorder

7 (31.8%)

62 (80.5%)

χ2 (1) = 31.36 ***


Number of Hospitalizations

0.40 (0.84)

0.69 (1.21)

t (96) = −1.37

Cohen’s d = 0.23

Number of Emergency Room Visits

0.51 (0.86)

1.18 (1.19)

t (96) = −3.18**

Cohen’s d = 0.54

Suicide Attempt: (Dichotomous)

9 (22.5%)

62 (68.89%)

χ2 (1) = 18.90***


Number of Suicide Attempts

0.40 (0.84)

1.90 (2.91)

t (96) = −3.23**

Cohen’s d = 0.55

*p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001

Table 2

Correlation matrix and descriptive statistics













Emotion Regulation












2. EESC: Reluctance











3. Emotion words










Peer Relations


4. Overt Victimization









5. Relational Victimization








6. Negative Friendship Interactions







Family Characteristics


7. Family Cohesion






8. Family Conflict





9. Punish-C






10. Frequency



11.. Methods


Mean (SD)

105.10 (23.54)

23.92 (6.92)

10.90 (6.75)

5.09 (2.41)

11.10 (4.72)

1.66 (0.71)

4.52 (2.76)

4.67 (2.61)

2.36 (0.53)

2.43 (1.91)

2.04 (1.61)













The Hypothesized Model

The hypothesized model is illustrated in Fig. 1. Circles represent latent variables, and rectangles represent measured variables. Absence of a line connecting variables indicates lack of a hypothesized direct effect. The hypothesized model examined the predictors of NSSI behavior. NSSI was a latent variable with two indicators: the frequency of NSSI behavior and the number of methods used to engage in NSSI behavior. It was hypothesized that emotion dysregulation abilities (a latent variable with three indicators including DERS emotion dysregulation patterns, EESC expressive reluctance, and LIWC lack of emotional awareness as measured by the emotional reflection task [number of Emotion Words]), family relational problems (a latent variable with three indicators including EAC socialization strategies, FES family conflict, and lack of family cohesion), and peer relational problems (a latent variable with three indicators including SEQ peer relational aggression, SEQ peer overt aggression, and NRI negative friendship interactions) directly predicted NSSI behavior. Additionally, it was hypothesized that ER would partially mediate the relationship between interpersonal difficulties (family and peer) and adolescents’ NSSI behavior.
Fig. 1

Proposed SEM model of interpersonal difficulties influencing NSSI as mediated by emotion dysregulation

Model Fit

Model fit was evaluated based on three indicators including the Root Mean Square Error of Approximation (RMSEA), Normed Fit Index (NFI), and Comparative Fit Index (CFI). RMSEA fit indices of zero are considered a perfect fit, and values less than 0.05 are considered a close fit. CFI and NFI values range from zero to one with one representing a perfect fit. Values above 0.90 are considered to be excellent (Tabachnick & Fidell 2001). The results based on the hypothesized model indicated a strong fit, χ2 (39, N = 99) = 40.72, p = 0.40, RMSEA = 0.02, NFI = 0.87, CFI = 0.99. Despite this strong fit, post-hoc model modifications were performed to develop a better fitting model. Alternative models were investigated using nested and non-nested models.

Nested Models

Model trimming was conducted through the evaluation of a chi-square difference test. In the first model, the direct and mediated effects of family relational problems on peer relational problems and emotion dsyregulation on NSSI were included and the model was fully unconstrained. In the second model, the path for the direct effect of family relational problems on peer relational problems was set to zero. In the third model, the path for family relational problems on NSSI was set to zero. In the fourth model, the path for peer relational problems on NSSI was set to zero. In the fifth model, the direct effects from peer and family relational problems were set to zero. Evaluation of the chi-square difference revealed no significant difference between the unconstrained original model, χ2 (39, N = 99) = 40.72, p = 0.40 and the nested constrained modified model, Δχ2 (40, N = 99) for the path of peer relational problems on NSSI, Δχ2 (1) = 1.04, p = 0.31. All other constrained modified models remained significant. Thus, the deletion of the path from peer relational problems to NSSI was accepted based on grounds of parsimony. The final model fit the data well, χ2 (40, N = 99) = 41.76, RMSEA = 0.02, HFI = 0.86, CFI = 0.99. The model with a direct effect of family relational problems on peer relational problems, emotion dysregulation, and NSSI as well as indirect effects of peer and family relational problems on NSSI through emotion dysregulation is presented in Fig. 2.
Fig. 2

Final SEM model of interpersonal difficulties influencing NSSI as mediated by emotion dysregulation. Note. ** p ≤ 0.01, * p ≤ 0.05, ^ p = 0.08

Non-Nested Models

A plausible alternative model that examined the direction of effects issue was evaluated against the final model identified in Fig. 2. The alternative model (Fig. 3) tested family and peer relational problems as mediating the relation between girls’ emotion dysregulation and NSSI behaviors. The alternative model was determined to not be a good fit, χ2 (48, N = 99) = 381.28, p < 0.01. The AIC criteria were compared in the final (AIC = 111.76) and alterative model (AIC = 366.89), indicating that the final model was a better fit. The final model with standardized coefficients is presented in Fig. 2.
Fig. 3

Alternative SEM model of emotion dysregulation influencing behaviors as mediated by interpersonal difficulties. Note. ** p ≤ 0.01, * p ≤ 0.05, ^ p = 0.08

Direct Effects

Direct effects in the final model (Fig. 2) indicate increased NSSI behavior was predicted by greater emotional dysregulation (unstandardized coefficient = 0.13, standardized coefficient = 0.41, p = 0.01). The direct relationship between family relational problems and increased NSSI was marginally significant (unstandardized coefficient=4.97, standardized coefficient = 0.38, p = 0.08). The direct effect of peer relational problems on increased emotion dysregulation was also marginally significant (unstandardized coefficient=7.41, standardized coefficient = 0.34, p = 0.06).

Indirect Effects

The significance of the intervening variables in the final model (Fig. 2) was evaluated using tests of indirect effects through AMOS. This model of examining intervening variables has more power than the mediating variable approach (MacKinnon et al. 2002).

Girls’ emotion dysregulation served as an intervening variable for family and peer interpersonal difficulties. Greater family relational problems predicted emotion dysregulation, which in turn predicted NSSI behaviors (unstandardized indirect effect=2.76, standardized indirect effect=0.09, p = 0.01). Similarly, more peer relational problems predicted girls’ emotional dysregulation, which in turn predicted NSSI behaviors (unstandardized indirect effect = 0.98, standardized indirect effect = 0.11, p = 0.005).


Overall, this research provides support for the hypothesis that the linkage between emotional dysregulation and NSSI is fostered through unsupportive social contexts (Crowell et al. 2009). Structural equation modeling revealed an adequate fit for the model that included direct effects of family relational problems and the indirect effects of family and peer relational problems through emotion dysregulation for adolescent girls who were psychiatrically hospitalized. Further, this study adds support to the conceptualization of emotion dysregulation as a core feature of NSSI. As such, the results indicate that family and peer interpersonal problems have negative and independent effects on emotional dysregulation, and dysregulated emotional states have significant influence on NSSI. These findings extend research that has demonstrated children’s regulation of emotion develops within a social context and is related to psychological and social adjustment (Zeman et al. 2006).

Direct and Indirect Effects of Family Relational Problems

The results point to the direct role of family relational difficulties on NSSI and the indirect role through developing inadequate ER patterns and indicate that these roles may be important risk factors for self-injury. The finding regarding family relational problems, however, should be viewed with caution given its marginal significance. These results are nevertheless consistent with research that has demonstrated that families of clinic-referred, self-injuring adolescent girls exhibit less positive affect, more negative affect, lower family cohesiveness, and higher rates of conflict/negativity when compared to non-psychiatric controls during an interaction task requiring discussion of a contentious current issue (Crowell et al. 2008). Moreover, the broader literature on suicidal behaviors and family conflict has suggested similar patterns of familial conflict that place youth at risk (e.g., Wagner 1997).

The model indicated that family relational problems also presented significant direct risk in predicting poor peer experiences. Prior research has illustrated important family and peer linkages, in which developing peer social competence has its roots in familial relationships through direct and indirect mechanisms (Carson and Parke 1996). Familial relational problems may interfere with the development of an internal locus of control and healthy self-concept that in turn places children and adolescents at risk for peer harassment and victimization (Barber 1996). Moreover, the coercive relational patterns and punitive disciplinary practices that accompany many forms of familial relationship problems predict antisocial behavior and poor adjustment within the peer context (Dishion 1990).

Support was obtained for the hypothesis that family relational problems would affect NSSI through emotional dysregulation. This finding is consistent with the emotion development literature, which suggests that parental responses to their children’s emotional displays have significant effects on the subsequent awareness, expression, regulation, and coping with emotion (Gottman et al. 1997). Further, Gottman and colleagues emphasize the importance of parental validation of emotion as a primary factor in determining children’s subsequent psychosocial functioning. That is, parental invalidation is associated with socio-emotional problems and difficulties with down-regulating emotional arousal. The present study indicated that adolescents’ ER abilities partially mediated the relationship between interpersonal difficulties in the family and NSSI. In other words, adolescents with difficulties identifying and expressing their negative emotions within an unsupportive environment were less equipped to manage strong negative emotional experiences in adaptive ways. Moreover, the direct relationship between family and peer interpersonal difficulties suggests that being exposed to such problems also affects other emotional competence skills such as decoding others’ emotions, developing empathy, and emotion perspective taking.

The indirect relation found between family relational problems and NSSI in this study is consistent with the extant literature that suggests the relation between family factors and suicidal behaviors is through a causal chain process in which exposure to childhood adversity increases youths’ vulnerability to psychopathology and stress, which in turn leads to increased risk of suicidal behaviors (Gratz et al. 2002). Previous research indicates that high conflict and hostility in families is associated with the under control of emotion in their children (Rothbaum and Weisz 1994). Although this study measured parental socialization retrospectively, it suggests that adolescents’ perceptions of childhood interpersonal difficulties contribute to patterns of dysregulated emotional experiences with long term consequences.

Indirect Effects of Peer Relational Problems

The results of the present study reveal that peer processes also play an important role in NSSI. Specifically, interpersonal difficulties experienced within the peer context lead to NSSI behavior due to their effect on emotional dysregulation. The role of peers as being essential in maintaining normative social development is a well-established tenet in the developmental literature (Parker and Asher 1987). This study adds to the literature by illuminating processes that occur within adolescents’ peer relationships that are central to the continued fine-tuning of ER skills and adaptive functioning in later adolescence. The effects of peer relational problems were mediated through emotional dysregulation, suggesting that peers have a powerful influence on how adolescents manage emotion and subsequent behavior. These findings dovetail those of Hilt et al. (2008) who examined interpersonal distress and its relation to NSSI in a community sample of young adolescent girls. The authors reported specific relations between social experiences and the functions and goals of NSSI. For example, adolescent girls with low quality peer communication and high peer victimization were the most likely to report the social reinforcement function of NSSI.

Although the present data are limited to describing the peer functioning factors that are specifically predictive of NSSI, research on precipitants of suicidal behavior indicates that victimization by peers is a common risk factor (Prinstein 2003). Moreover, several investigations have revealed significant prospective pathways in which a lack of peer support leads to increased loneliness, depression, and internal distress (e.g., Nangle et al. 2003). This internal distress, without the skills and support to manage intense emotional experiences, may lead to NSSI. Thus, negative peer experiences contribute to developmental vulnerability that leaves some adolescents insufficiently equipped to respond to emotionally intense experiences in adaptive, constructive ways. Consequently, they may choose to turn to NSSI as a means of managing their high emotional arousal that overwhelms and exceeds their coping resources.

Limitations and Future Directions

Although the results of the present study indicate that emotion dysregulation and family and peer interpersonal difficulties relate to NSSI, it should be noted that there are several limitations that may restrict the interpretation of the findings. First, the use of a cross-sectional design limits the conclusions that can be drawn about the role of family and peer interpersonal difficulties in the development and maintenance of NSSI. Although structural equation modeling did not support an alternative direction of effects (i.e., NSSI leading to family and peer interpersonal difficulties), longitudinal modeling is needed to clarify these questions. Second, characteristics of the sample, including the use of a psychiatrically hospitalized sample of adolescent girls who were predominately middle class and Caucasian, limit the generalizability of the results. Third, although the focus of this study was not on examining NSSI within diagnostic phenotypes, a systematic evaluation of psychiatric status would have further strengthened the description of this sample. Fourth, the literature about NSSI behavior in adolescent boys is strikingly absent and clearly is an avenue for future examination. It is unclear whether the current model would hold for boys given that the nature of boys’ and girls’ peer relations generally differs (e.g., boys’ friendships tend to be less intimate; Buhrmester and Furman 1987). Fifth, this study relied exclusively on self-report methods. This was necessary, given that the constructs of interest involved individuals’ perceptions of their behavior and of their family and peer experiences, which are assessed most effectively in adolescents through self-report (Inderbitzen 1994). However, future work should include other reporting sources to ensure the results do not solely reflect shared method variance.

Despite its limitations, this investigation of adolescent NSSI contributes to the literature in important, novel ways. Specifically, this study presents an integrated, theoretically-based model of risk factors that play an important role in NSSI in adolescent girls. The study utilized modeling to illustrate the complex relationships between an individual’s environment and her adaptation. The findings provide a significant advance in the understanding of how NSSI behavior may operate through a cycle of difficult relationships that contribute to intense emotional arousal combined with a lack of skills or support to manage those experiences. Under these circumstances, it appears that adolescent girls who are under distress and do not have adaptive coping mechanisms are at heightened risk to turn to NSSI.


This publication was made possible by grant number T32 DH052462 from the National Institute of Child Health and Human Development (NICHD), NIH and the Mayo Department of Psychiatry and Psychology, 2007 Small Grants Award.

Copyright information

© Springer Science+Business Media, LLC 2010