Abstract
This study examined the individual functioning, interpersonal relations, and academic performance of 142 female adolescents who were divided into three demographically similar groups: violent juvenile offenders (n = 51), nonviolent juvenile offenders (n = 56), and nondelinquent youths (ND; n = 35). Both the violent and the nonviolent offenders had histories of serious crimes (i.e., felonies). Parents and youths completed self-report instruments, behavior-rating inventories, and a videorecorded interaction task, and teachers completed a rating measure. Results showed that violent juvenile offenders and nonviolent juvenile offenders had more behavior problems, more difficulties in family and peer relations, and poorer academic performance than did ND. However, violent juvenile offenders did not differ from nonviolent juvenile offenders on any of the measures of individual functioning, family relations, peer relations, or academic performance. Taken together, the findings suggest that girls who commit serious crimes, regardless of whether violence is involved, share a number of risk factors across key domains of individual and interpersonal functioning. The implications of these findings for research, theory, and treatment are discussed.
Adolescent girls have been increasingly represented in the juvenile justice system over the past 15 years and now account for approximately 34% of all serious (i.e., index) crimes committed by youths (Federal Bureau of Investigation, 1995, 2010). In response to this trend, federal study groups (e.g., Office of Juvenile Justice and Delinquency Prevention, 2010) and reviewers (e.g., Sprott & Doob, 2009; Zahn, 2009) have recently attempted to advance our understanding of the causes and correlates of criminal behavior among adolescent girls so that effective prevention and treatment programs can be developed. Indeed, when one considers that girls who commit serious violent and nonviolent crimes during adolescence are at high risk of mental and physical health problems, substance abuse, low educational and vocational achievement, and persistent criminal behavior in adulthood (Bright & Jonson-Reid, 2010; Piquero, Brame, & Moffitt, 2005), it seems imperative to develop an empirical base regarding the characteristics of female serious juvenile offenders and the systems in which they are embedded.
Unfortunately, however, very few controlled studies have examined the characteristics of female serious violent offenders (VO) and nonviolent offenders (NO), and the vast majority of studies have included relatively significant methodological limitations (Goldweber, Broidy, & Cauffman, 2009; Wong, Slotboom, & Bijleveld, 2010). For example, research on female VO has typically not used appropriate comparison groups (e.g., NO with similar demographic characteristics). Without such control groups, it is difficult to determine whether observed results are linked with violent offending in particular or with delinquency in general. Furthermore, the data in many of the extant studies are derived from clinical impressions and unstandardized assessment instruments, and the self-reports of incarcerated adolescents are often the primary source of information. In light of these methodological difficulties, findings from research conducted to date should be viewed as quite tentative.
In developing an understanding of female serious juvenile offenders, it seems most efficacious to examine variables that are consistent with comprehensive theories of delinquency. Agnew (2009) has validated a general integrated theory of crime and delinquency that combines elements of strain theory, control theory, and social learning theory to explain delinquency in both girls and boys. This integrated theory proposes that low bonding to conventional socializing agents (i.e., family, school) increases the likelihood that youths will associate with deviant peers, and association with such peers is viewed as the primary determinant of delinquent behavior. Similarly, developmentalists (e.g., Bronfenbrenner & Morris, 2006; Minuchin, 2002) and family systems theorists (e.g., Henggeler, Schoenwald, Borduin, Rowland, & Cunningham, 2009; Minuchin & Nichols, 1998) have emphasized that child behavior is associated with the bidirectional interaction between individual child characteristics and the key systems or contexts in which youths are embedded. Thus, both integrated theory and contextual/systemic approaches highlight the roles of individual child characteristics, family relations, and extrafamilial factors (e.g., peers, school) in the development of antisocial behavior.
Integrated theory and contextual/systemic perspectives have stressed the importance of family bonding (i.e., affective ties) and family organization (i.e., adaptability, control) in the development of delinquency. Indeed, investigators have found that female delinquent youths report high rates of conflict, disorganization, and dysfunction and low rates of positive communication in their families (Dixon, Howie, & Starling, 2004; Fergusson & Woodward, 2000; Hubbard & Pratt, 2002). However, studies have focused primarily on girls who commit nonviolent offenses and have not examined whether such family problems are associated with violent offenses among girls. It is possible that girls who commit violent crimes, like their violent male counterparts (see Blaske, Borduin, Henggeler, & Mann, 1989), have even more dysfunctional family relations than do NO.
A central aspect of the integrated model is that delinquent youths associate extensively with deviant peers. In contrast with social learning theory (e.g., Snyder, Reid, & Patterson, 2003), however, the integrated model does not suggest that the peer relations of delinquent girls are deficient in positive affect and social skills (Agnew, 2009). Instead, studies using community samples have shown that involvement with deviant peers is associated with increased delinquency among girls (Fergusson & Woodward, 2000; Hubbard & Pratt, 2002). Although there is evidence that involvement with deviant peers is more strongly associated with violent than with nonviolent criminal behavior among boys (Blaske et al., 1989; Lacourse, Nagin, Tremblay, Vitaro, & Claes, 2003), it is not clear whether female serious VO associate more extensively with deviant peers than do female serious NO.
The integrated and contextual/systemic models also emphasize the role of school bonding (i.e., academic performance, involvement in extracurricular activities) in the development of delinquency. In general, investigators have reported that delinquent girls demonstrate poorer academic performance than do nondelinquent girls (Hubbard & Pratt, 2002; Payne, 2009). Although research has shown that boys who commit violent crimes have more academic difficulties than do boys who commit nonviolent crimes (Lipsey & Derzon, 1998; Stouthamer-Loeber & Loeber, 2002), studies have not compared female VO and NO on measures of school bonding.
Finally, from both integrated and contextual/systemic perspectives, individual youth characteristics are also important correlates of delinquent behavior. The extant research with delinquent girls has emphasized the assessment of psychiatric symptomatology rather than the study of cognitive and social developmental processes. This literature suggests that female delinquent youths report high rates of emotional and behavioral problems (Dixon et al., 2004; Hubbard & Pratt, 2002). Again, however, studies have not focused on girls who commit serious crimes and have not included pertinent comparison groups.
In summary, investigators have concluded that female delinquent youths generally evidence emotional and interpersonal deficits that are consistent with major theories of crime and antisocial behavior. However, our understanding of the characteristics of female serious juvenile offenders has been limited by a relative absence of controlled studies and by other methodological problems in the extant literature. The primary purpose of the present study is to provide a rigorous evaluation of the characteristics of female serious juvenile offenders and of the key systems in which they are embedded. As such, this study includes several relative strengths. First, consistent with previous research indicating that male violent and nonviolent juvenile offenders represent distinct subgroups (see Loeber & Farrington, 1998), the present study distinguishes between girls who have committed serious violent crimes and girls who have committed serious nonviolent (i.e., property) crimes. Second, we include a demographically similar comparison group of nondelinquent girls to help determine the unique and common correlates of different forms of female offending. Finally, the measurement methodology includes both standardized self-report measures that are obtained from multiple respondents and observational measures of family interaction. As described subsequently, we selected measures that reflected the central constructs used in the integrated and systemic theories regarding the causes of delinquent behavior.
Method
Participants
One hundred and forty-two female participants and their parent/parents were involved in the study. Participants ranged in age from 11 to 17 years and were divided into three groups: VO (n = 51), NO (n = 56), and nondelinquent youths (ND; n = 35). The two groups of female offenders were a subset of a larger sample (N = 599) of adjudicated delinquent youths who had been referred consecutively from the local juvenile office to the Missouri Delinquency Project (Schaeffer & Borduin, 2005). Referrals to the larger sample included all families in which the youth (a) had at least two arrests, (b) had been detained previously for at least 4 weeks, (c) was currently living with at least one parent figure, and (d) showed no evidence of psychosis or dementia. The selection criteria for the VO group included at least one arrest for a felony violent crime (i.e., aggravated assault, robbery). Consistent with the literature on violent and nonviolent male juvenile offenders (see Loeber & Farrington, 1998), all of the girls in the VO group had also been arrested for felony nonviolent crimes and had more arrests overall (M = 7.33, SD = 4.35), including misdemeanor violent crimes, than did girls in the NO group, M = 4.75, SD = 2.93; Welch’s F(1, 24.96) = 5.17, p = .032. In our larger sample, no girls had been arrested only for misdemeanor violent crimes. The NO group had (a) at least one arrest for a felony nonviolent crime (i.e., theft, burglary, larceny) and (b) no arrests for felony or misdemeanor violent crimes.
The ND youths were recruited from registries of public school students during the same time that juvenile offenders were referred to the project. The ND youths and their families were from the same neighborhoods as the families in the offender groups but had no history of arrest or inpatient psychiatric treatment. Thus, the ND youths did not evidence serious problems; yet, they were from similar sociocultural backgrounds to those of the juvenile offenders. In addition, the ND girls’ scores on Quay and Peterson’s (1987) Revised Behavior Problem Checklist (RBPC) were similar to the norms for nonclinic female adolescents.
Primary caregivers in the present study ranged in age from 30 to 66 years (M = 39.5, SD = 6.04) and were predominantly biological mothers (n = 128), although the sample also included four stepmothers, three adoptive mothers, two biological grandmothers, four biological fathers, and one adoptive father as primary caregivers. As shown in Table 1, the VO, NO, and ND groups did not differ significantly in youth age (M = 14.4 years, SD = 1.38), grade in school (44.4% were in the eighth or ninth grades), social class (34.5% were lower socioeconomic status, corresponding to Class IV or V; Hollingshead, 1975), race (80.3% were Caucasian, 19.0% were African American, and 0.7% were other), or single-parent households (53% across groups). In addition, the two groups of offenders did not differ from each other on age at first arrest (M = 13.0 years, SD = 1.78).
Participant Background Characteristics by Group
Note. VO = violent offenders; NO = nonviolent offenders; ND = nondelinquent youths. Means that do not share any common subscripts across rows are significantly different at the .05 level. For youth age, df = 2, 75.99; for youth grade in school, df = 2, 72.86; for age of first arrest, df = 1, 35.63; for social class, df = 8; for race, df = 4; for single-parent households, df = 2.
aBased on Hollingshead’s (1975) Four-Factor Index of Social Status.
Procedure
The interviewers were graduate and advanced undergraduate students in psychology who were not aware of the purpose of the study and did not have access to clinical information regarding the families. Each interviewer received approximately 20 hr of training prior to the first family contact to standardize the assessment procedures and to recognize and attenuate circumstances (e.g., fatigue, reading problems) that threatened the validity of the assessments.
All families were initially contacted by telephone or by a home visit and were asked to participate in a study of youth individual adjustment and interpersonal relations. Families of juvenile offenders were told that participation in the study was voluntary and that refusal to participate would not jeopardize their receipt of court services (e.g., psychotherapy, academic tutoring). The youths remained under the jurisdiction of the court regardless of their families’ decisions about participating in the study. Approximately 90% of the families that were contacted agreed to participate in the study. All procedures were approved by the Institutional Review Board of the University of Missouri.
The assessment session took place at the family’s convenience either in their home or in a youth center in their neighborhood. Across groups, the majority (87%) of the families completed the assessment in their homes. In two-parent households, both of the parents participated in the assessment. At the beginning of each session, the interviewer reviewed the general procedure and purpose of the assessment, explained confidentiality and consent issues, and obtained written consent from the parent/parents and written assent from the youth. The interviewer then administered a series of self-report (i.e., pencil and paper) instruments and behavior-rating inventories in a random order to the parent/parents and youth. Each of these measures was filled out privately and without discussion. In addition, the parent/parents and youth were videorecorded as they discussed and jointly completed a family interaction task. Family members were instructed to work at their own pace and to inform the interviewer when they had completed the task. The interviewer then placed a pencil and blank questionnaire on the table next to the family and started a videorecorder. The interviewer left the room until the discussion was completed (M discussion time = 10.24 min).
One of the youth’s teachers also completed a paper-and-pencil instrument. The teacher was randomly selected from a list of the youth’s current teachers. The teacher was told that the youth was a participant in a study of adolescent socialization.
Dependent Measures
Individual adjustment
As described earlier, both the integrated and the contextual/systemic perspectives have emphasized that individual characteristics of youths are important correlates of delinquent behavior. Consistent with extant research, individual characteristics were assessed in terms of youth behavioral and emotional problems.
RBPC
Youth behavior problems were assessed through caregiver reports on the 89-item RBPC (Quay & Peterson, 1987). The RBPC measures four dimensions of adolescent psychopathology: anxiety-withdrawal, attention problem, conduct disorder, and socialized-aggression. The subscales have strong psychometric properties (McMahon & Frick, 2007) and have discriminated between delinquent and nondelinquent adolescents (e.g., Armistead, Wierson, Forehand, & Frame, 1992). In the present study, internal consistency coefficients (αs) for the RBPC subscales ranged from .87 to .95.
Brief Symptom Inventory (BSI)
Self-reported youth psychological distress was assessed by the Global Severity Index (GSI) of the 53-item BSI (Derogatis, 1993). The GSI is considered the best single index of respondent emotional distress and is formed by summing scores across the items and dividing by the total number of items. Coefficient α for the GSI in this study was .97. Research has supported the convergent and discriminant validity of the BSI with delinquent and nondelinquent males and females (e.g., Handal, Gist, Gilner, & Searight, 1993).
Family relations
In this study, we chose measures of cohesion and warmth to represent central aspects of family bonding. We used measures of adaptability and conflict to assess the level of organization in the family. Family bonding (i.e., cohesion, warmth) and family organization (i.e., adaptability, control) represent important explanatory constructs in the contextual/systemic and integrated theories of delinquent behavior. The use of both self-report and observational methods provided different vantage points on family transactional patterns.
Family Adaptability and Cohesion Evaluation Scales-II (FACES-II)
Parent and youth perceptions of family relations were evaluated with the 30-item FACES-II (Olson, Portner, & Bell, 1982). The FACES-II assesses the dimensions of cohesion, which refers to the emotional bonding and individual autonomy of family members, and adaptability, defined as the capacity of the family system to change its power structure, role relations, and relationship rules in response to situational and developmental stress. Coefficient αs for the cohesion and adaptability subscales, respectively, were .88 and .83 for youths’ reports, and .90 and .78 for parents’ reports in the present study. The subscales have proven useful in the study of general delinquency (Matherne & Thomas, 2001; Rodick, Henggeler, & Hanson, 1986). Consistent with Olson’s (2000) recommendations, we treated adaptability and cohesion as linear scales in subsequent statistical analyses. In two-parent households, we averaged the ratings of the parents to create composite scores of adaptability and cohesion.
Observational measures
Observational measures were based on the family members’ videorecorded discussion on the Unrevealed Differences Questionnaire-Revised (Borduin et al., 1995), which consists of 9 items pertaining to emotional issues, decision making, and discipline. Each item includes three to six alternative choices, and the family was instructed to rank the choices in order of preference.
The observational measures included ratings of warmth and conflict, which were completed after the observers had watched the entire videotaped discussion. Affect was rated on a scale from 1 (very cold) to 7 (very warm), and conflict was rated on a scale from 1 (none) to 5 (extreme). These observational measures have been used in prior research with families of juvenile offenders (e.g., Borduin et al., 1995; Ronis & Borduin, 2007).
Raters included graduate and advanced undergraduate students who received 20 hr of training prior to scoring the videotaped family interactions. Interrater reliability was assessed throughout the study and was determined on 26% of the families. Cohen’s κ values were .85 for the bipolar rating of warmth and .89 for the bipolar rating of conflict.
Peer relations
We examined several aspects of youth peer relations that are pertinent to comprehensive theories of delinquent behavior. Association with deviant peers, a central aspect of the integrated model, was assessed using the socialized-aggression subscale of the RBPC. In addition, emotional bonding to peers, which occupies a central role in control theory formulations of delinquent behavior, was measured using the Missouri Peer Relations Inventory (MPRI; Borduin, Blaske, Cone, & Mann, 1989). Moreover, family systems theory and developmental contextual approaches emphasize that adolescence is accompanied by increased autonomy from parents, stronger bonding to peers, and more susceptibility to antisocial peer influences (Jaccard, Blanton, & Dodge, 2005).
We obtained youth, parent, and teacher ratings of the youth’s peer relations with the 13-item MPRI. To define the main dimensions that were tapped by the MPRI, we conducted an exploratory factor analysis on respondents’ ratings. As recommended by Comrey (1988) and Patil, McPherson, and Friesner (2010), the number of factors to be extracted in the solution was determined through scree plot analysis and parallel analysis (using computer-generated random data) and was fixed. We used maximum likelihood extraction with Promax rotation, and only those variables with factor loadings on the pattern matrix greater than .4 were considered to define a factor. Two factors, accounting for approximately 45% of the variance, emerged from this analysis. The first factor, emotional bonding, included items that reflect emotional warmth and closeness with peers. The second factor, aggression, included items that assess an aggressive, acting-out style of peer interaction. The construct validity of these dimensions has been supported in other studies of serious juvenile offenders (e.g., Borduin et al., 1989; Ronis & Borduin, 2007). In the current study, coefficient αs for emotional bonding and aggression were .77 and .75, respectively.
For subsequent analyses, factor scores were derived from the variables that tapped each dimension of peer relations. We computed each factor score by multiplying the standardized score for each variable by the appropriate factor score coefficient and then summing the resulting products over all variables in the factor (see Grice, 2001).
Academic performance
Both integrated and contextual/systemic perspectives have emphasized that low bonding to school is associated with delinquent behavior. Consistent with these perspectives, we assessed youth academic performance as an index of school bonding.
Four items were used to assess youth academic performance (i.e., grades) in respective content areas: math, English, social studies, and science. Youths, parents, and teachers reported the youth’s grade in each area using a 5-point scale (ranging from 0 = grade of F to 4 = grade of A). Internal consistency (α) for a composite grade score (averaged across respondents and content areas) was .92.
Results
The three groups were compared on measures of individual adjustment, family relations, peer relations, and academic performance. In light of the relatively low sample size, we used a series of one-way univariate analyses of variance (ANOVAs) rather than multivariate ANOVAs followed by univariate analyses (for a discussion of this issue, see Huberty & Morris, 1989; Huizenga, Smeding, Grasman, & Schmand, 2007). When a significant ANOVA was revealed, multiple pairwise comparisons were completed to determine the nature of the between-groups differences. Following the recommendations of Jaccard and Guilamo-Ramos (2002), we used a Bonferroni-adjusted α level in the pairwise comparisons to minimize familywise error. Familywise error was also reduced through the use of data aggregation (i.e., factor scores, composite scores).
Separate effect size indexes, reported as eta-squared (η2) values, were calculated for each ANOVA and are included in the respective tables. Based on Cohen’s (1988) criteria, the majority of the effect sizes were “medium” (η2 = .06) to “large” (η2 = .14).
Individual Adjustment
RBPC
As shown in Table 2, univariate analyses and post hoc comparisons revealed that parents in the VO and NO groups reported more anxiety-withdrawal, attention problems, conduct problems, and socialized-aggression by their daughters than did parents in the ND group. Youths in the VO and NO groups did not differ in their behavior problems.
Group Means and Standard Deviations for Measures of Youth Individual Adjustment
Note. VO = violent offenders; NO = nonviolent offenders; ND = nondelinquent youths; RBPC = Revised Behavior Problem Checklist; BSI (GSI) = Brief Symptom Inventory (Global Severity Index). Means that do not share any common subscripts across rows are significantly different at the .017 (.05/3) level. For anxiety-withdrawal, df = 2, 85.28; for attention problem, df = 2, 84.96; for conduct disorder, df = 2, 86.37; for socialized-aggression, df = 2, 71.39; for the BSI (GSI), df = 2, 76.82.
aIn two-parent households, parents’ reports were averaged.
bYouth report.
**p < .01.
BSI
Based on the ANOVA, there were no between-groups differences on youth reports of psychological distress (see Table 2).
Family Relations
FACES-II
As presented in Table 3, univariate analyses and post hoc comparisons revealed that parents and youths in the VO and NO groups reported lower levels of family cohesion and adaptability than did their counterparts in the ND group. The overall effect sizes were medium to large. There were no differences between the VO and NO groups on parent or youth reports of cohesion or adaptability.
Group Means and Standard Deviations for Measures of Family Relations
Note. VO = violent offenders; NO = nonviolent offenders; ND = nondelinquent youths; FACES-II = Family Adaptability and Cohesion Evaluation Scales-II. Means that do not share any common subscripts across rows are significantly different at the .017 (.05/3) level. For parent cohesion, df = 2, 128; for parent adaptability, df = 2, 83.01; for youth cohesion, df = 2, 129; for youth adaptability, df = 2, 84.79; for observational warmth, df = 2, 106; for observational conflict, df = 2, 107.
aIn two-parent households, parents’ reports were averaged.
**p < .01.
Observational measures
A significant univariate effect emerged for the observational measure of warmth (see Table 3). Post hoc comparisons revealed that families in the NO group exhibited less warmth than did families in the ND group; however, there were no significant differences between the VO and the NO or ND groups. A significant univariate effect did not emerge for the observational measure of conflict.
Peer Relations
As shown in Table 4, univariate and post hoc analyses indicated that VO and NO youths were rated as lower in emotional bonding with peers and as higher in aggression toward peers than were ND youths. The overall effect sizes for these measures were medium to large. There were no significant differences between the VO and NO groups in ratings of their peer relations.
Group Means and Standard Deviations for Measures of Youth Peer Relations and Academic Performance
Note. VO = violent offenders; NO = nonviolent offenders; ND = nondelinquent youths; MPRI = Missouri Peer Relations Inventory. Means that do not share any common subscripts across rows are significantly different at the .017 (.05/3) level. For emotional bonding, aggression, and academic performance, dfs = 2, 134.
aIn two-parent households, parents’ reports were averaged.
**p < .01.
Academic Performance
Univariate and post hoc analyses revealed that VO and NO youths had lower grades in school than did ND youths (see Table 4). The VO and NO groups did not differ in their grades.
Discussion
In this study, we examined the characteristics of girls with histories of serious criminal behavior. Our study was designed to improve on prior work by including two groups of female serious juvenile offenders differentiated by type of crime (i.e., violent vs. nonviolent), a demographically similar comparison group of nondelinquent girls, and standardized measures obtained from multiple perspectives. Our results suggest that girls who commit serious crimes, regardless of whether violence is involved, have a number of common problems. Both groups of female juvenile offenders evidenced disturbances in their family and peer relations that are consistent with integrated theory (Agnew, 2009) and support the role of differential affective ties to conventional and deviant socializing agents. Regarding conventional socializing agents, female VO were generally similar to NO and showed lower bonding to family (FACES-II, cohesion subscale), school (grades), and prosocial peers (MPRI, aggression and emotional bonding subscales) than did ND youths. Regarding deviant socializing agents, female VO were like NO in their frequent association with delinquent peers (RBPC, socialized-aggression subscale). Thus, in general, the results suggest that girls who commit serious violent or nonviolent crimes have low bonding to family and school and high involvement with deviant peers. These results contribute to the small empirical base regarding female serious juvenile offenders, whose high recidivism rates are of great concern to policy makers (see Davis, Fisher, Gershenson, Grudzinskas, & Banks, 2009; Johansson & Kempf-Leonard, 2009),
Consistent with the “common features” model of child psychopathology (see Lilienfeld, 2003; Windle & Mason, 2004), the two groups of female serious offenders in the present study displayed a number of common problems but no problems that were specific to a particular pattern of offending. It is possible that, in contrast to serious violent and nonviolent offending in males, the same causal factors underlie serious violent and nonviolent offending in females. Indeed, research indicates that adverse life events, especially physical or sexual abuse, are stronger predictors of serious offending in girls than boys (Ariga et al., 2008; Dixon, Howie, & Starling, 2005; Kempf-Leonard & Johansson, 2007). In the present study, 72.5% of the VO and 57.1% of the NO had documented histories of physical or sexual abuse (determined through court records) by natural parents or other caregivers, and these percentages are likely underestimates given that most instances of abuse go unreported (Gilbert et al., 2009). It is possible that abuse by caregivers plays a causal role in girls’ withdrawal from key conventional bonds (e.g., parents, school) that would otherwise deter involvement in serious delinquent behavior. Moreover, there is evidence that girls who are victims of abuse often feel powerless in their interpersonal relations and frequently engage in either violent behavior (e.g., physical assault) to enhance their status among peers (Chesney-Lind & Belknap, 2004) or nonviolent behavior (e.g., burglary) to obtain tangible possessions (Lansford et al., 2007). Therefore, victimization may be either an indirect or direct cause of both serious violent and nonviolent criminal acts perpetrated by female youths.
It is also possible that female serious VO and NO have a number of problems in common as a result of being involved in the juvenile justice system. Indeed, regardless of the nature of their offenses, delinquent girls are often incarcerated in detention or residential facilities (Cauffman, Lexcen, Goldweber, Shulman, & Grisso, 2007; Chesney-Lind, Morash, & Stevens, 2008), which may expand their deviant peer networks and provide increased support for their criminal behavior (Bernburg, Krohn, & Rivera, 2006; Dodge, Dishion, & Lansford, 2006). In addition, involvement in the juvenile justice system may exacerbate girls’ mental health problems due to labeling (Bernburg et al., 2006) and reduced opportunities to be involved in prosocial activities (e.g., school jazz band, organized sports; Mason, Hitchings, McMahon, & Spoth, 2007). Furthermore, the juvenile justice system may increase stress and conflict in girls’ families (e.g., by requiring parents to attend hearings, pay court fees, and meet regularly with juvenile officers) and interfere with girls’ academic success (e.g., through prolonged exposure to low achieving peers both inside and outside of school). Nevertheless, to conclude with greater confidence that problems shared by female serious VO and NO are a result of their common experiences in the juvenile justice system, a longitudinal study is needed in which youths’ individual and interpersonal functioning is initially assessed prior to their history of arrests.
Several limitations of our study warrant comment. First, it is certainly possible that there are variables besides those examined in the present study that uniquely contribute to one type of serious offending or another. For example, there is evidence that hostile attributional biases (Dodge, 2003) and psychopathic traits (Cook, Barese, & DiCataldo, 2010) are relatively common among male violent juvenile offenders. It would be worth directly examining such variables among female violent juvenile offenders in a study with comparison groups like those in the present study. Second, our observational coding scheme assessed global dimensions of family interaction and did not assess transactional processes in family relations (see Kerig & Lindahl, 2001). For example, it seems important to examine the extent to which cross-generational coalitions (i.e., in which one parent forms a stable coalition with a child against the other parent) may play a role in the etiology or maintenance of adolescent girls’ criminal behavior (Buehler, Lange, & Franck, 2007; Mann, Borduin, Henggeler, & Blaske, 1990). Finally, it must be emphasized that our data do not address the directionality of causation for either serious violent or nonviolent criminal behavior in girls. For example, although integrated theory suggests that low family cohesion is an indirect cause of delinquent behavior, the low cohesion reported in families of VO and NO certainly could have resulted from the youths presenting serious and repeated behavior problems. It is also quite possible that low family cohesion and youth behavior problems are reciprocally related to each other (cf. Agnew, 2009). Nevertheless, although it is difficult to make causal determinations, it should be emphasized that the findings are consistent with key aspects of both the integrated model and contextual/systemic views of child psychopathology. Specifically, serious behavior problems in youths are best understood within the context of the key systems in which youths are embedded.
In conclusion, the results of this study suggest that girls who commit serious crimes, whether violent or nonviolent, share a number of risk factors across key domains of individual and interpersonal functioning. The broad nature of these risk factors has important implications for the design of interventions for girls who engage in serious crimes. The most commonly used interventions for female serious juvenile offenders (for a review, see Zahn, Day, Mihalic, & Tichavsky, 2009) are similar to those used for male serious juvenile offenders and typically address only a small subset of the possible determinants of delinquent behavior (e.g., poor anger management, low self-esteem). Our findings suggest that interventions for female serious juvenile offenders should address the broad range of problems that these youths show in their individual functioning (behavioral problems), family relations (lower cohesion and adaptability), peer relations (involvement with deviant peers, aggression toward peers), and academic performance (lower grades in school). Indeed, two broad-based treatments (i.e., multisystemic therapy [MST; Henggeler et al., 2009]; multidimensional treatment foster care [MTFC; Chamberlain, 2003]) have demonstrated considerable promise with this population of girls (Eyberg, Nelson, & Boggs, 2008; National Institutes of Health, 2006). Increased dissemination of these comprehensive treatments is consistent with our findings regarding the multiple problems experienced by female serious juvenile offenders.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.