Advertisement

Protection against antisocial behavior in children exposed to physically abusive discipline

      Abstract

      Purpose

      The study investigated protective factors (school commitment/importance, parent/peer disapproval of antisocial behavior, positive future orientation, and religion) hypothesized to lower risk for antisocial behavior among adolescents who, as children, had been physically abused. Protective factors also were investigated for comparison, nonabused children, and for children at risk on abuse and other factors: low socioeconomic status and early antisocial behavior.

      Methods

      Analyses used a two-step hierarchical regression approach. In step 1, age, gender, and early antisocial behavior were entered as controls. In step 2, each protective factor was entered separately as a predictor. A final regression model in each case examined the additive (combined) effect of all protective factors on a given outcome. Tests of predictor-by-group interactions were used to examine group differences.

      Results

      Among abused and nonabused children, having a strong commitment to school, having parents and peers who disapprove of antisocial behavior, and being involved in a religious community lowered rates of lifetime violence, delinquency, and status offenses. Having a positive future orientation appeared less powerful as a protective influence. Exposure to an increasing number of protective factors was for each outcome associated with a diminution in risk for antisocial behavior.

      Conclusions

      Protective factors represent targets for preventive intervention that are viable for children as they enter adolescence. The fact that protective factors were predictive of lower antisocial behavior in both the abuse and comparison groups suggests that protective effects are more universal than they are unique to a given group of children.

      Keywords

      Existing research has documented the deleterious effects of physical child abuse [
      • Widom C.S.
      Childhood victimization early adversity, later psychopathology.
      ]. For example, during adolescence, victims of abuse are at higher risk than are nonabused youth for a variety of mental health and behavior problems, including delinquency and violence [
      • Widom C.S.
      Childhood victimization early adversity, later psychopathology.
      ,
      • Fergusson D.M.
      • Lynskey M.T.
      Physical punishment/maltreatment during childhood and adjustment in young adulthood.
      ,
      • Herrenkohl R.C.
      • Egolf B.P.
      • Herrenkohl E.C.
      Preschool antecedents of adolescent assaultive behavior a longitudinal study.
      ,
      • Herrenkohl T.I.
      • Huang B.
      • Tajima E.A.
      • Whitney S.D.
      Examining the link between child abuse and youth violence an analysis of mediating mechanisms.
      ,
      • Smith C.
      • Thornberry T.P.
      The relationship between childhood maltreatment and adolescent involvement in delinquency.
      ,
      • Stouthamer-Loeber M.
      • Loeber R.
      • Homish D.
      • Wei E.
      Maltreatment of boys and the development of disruptive and delinquent behavior.
      ,
      • Tajima E.A.
      • Herrenkohl T.I.
      • Huang B.
      • Whitney S.D.
      Measuring child maltreatment a comparison of prospective parent reports and retrospective adolescent reports.
      ,
      • Widom C.S.
      The cycle of violence.
      ]. However, there is evidence that many victims of abuse avoid later involvement in antisocial behavior, which raises the prospect that these youth encounter protective influences that buffer the influence of this earlier risk exposure [
      • Widom C.S.
      Childhood victimization early adversity, later psychopathology.
      ]. Unfortunately, knowledge is weak on the range of protective influences that work against the onset of antisocial behavior in maltreated children. This study seeks to strengthen research on the topic of resilience by examining several protective factors hypothesized to lessen risk for various forms of antisocial behavior. The study also seeks to examine the degree to which the same protective factors are salient for a subsample of abused children who also meet criteria for risk on other factors (namely low socioeconomic status and early antisocial behavior), as well as children who comprise a “no-abuse” comparison group.

      Relation between child maltreatment and antisocial behavior

      Evidence on the relation between child maltreatment and youth antisocial behavior has emerged from a number of studies [
      • Widom C.S.
      Childhood victimization early adversity, later psychopathology.
      ]. In one such study, Stouthamer-Loeber et al. [
      • Stouthamer-Loeber M.
      • Loeber R.
      • Homish D.
      • Wei E.
      Maltreatment of boys and the development of disruptive and delinquent behavior.
      ] investigated the relation between child maltreatment (substantiated cases of abuse and neglect) and risk for overt (e.g., aggression, fighting, and violence) and covert (e.g., property damage and theft) forms of delinquency among boys from a longitudinal, inner city, community sample. They found that many boys who had been maltreated engaged in some later form of delinquency. Compared with controls (i.e., those in the sample without a history of abuse or neglect, matched on race, age, and socioeconomic status), a higher percentage of maltreated boys reported aggression (67% vs. 47%, respectively), fighting (77% vs. 43%, respectively), and serious physical violence (92% vs. 71%, respectively). Differences between maltreated boys and controls on covert forms of delinquency were less pronounced but still evident.
      Smith and Thornberry [
      • Smith C.
      • Thornberry T.P.
      The relationship between childhood maltreatment and adolescent involvement in delinquency.
      ] also investigated the relation between child maltreatment (before age 12; defined by official records) and adolescent delinquency. Data for their study were from the Rochester Youth Development Study, a multi-wave panel study of youth and their primary caregivers. After controlling for child gender, race, social class, family structure, and family mobility, they found a significant association between maltreatment and official delinquency; 45% of maltreated children engaged in later delinquency compared with 32% of those without maltreatment histories. A significant relation between maltreatment and youth self-reported delinquency also was revealed.
      In earlier analyses of data from the Lehigh Longitudinal Study, Herrenkohl et al. [
      • Herrenkohl R.C.
      • Egolf B.P.
      • Herrenkohl E.C.
      Preschool antecedents of adolescent assaultive behavior a longitudinal study.
      ] examined the relation between physical child abuse (defined by mothers’ reports of physically abusive discipline of a child in preschool) and youth-reported violent (assaultive) behavior in adolescence. Analyses showed that after accounting for demographic controls (SES, child gender, and age) and other forms of child maltreatment, mothers’ physically abusive discipline significantly predicted later violence.
      Widom and colleagues also have investigated the link between physical child abuse (officially recorded) and later violence in their well-known longitudinal cohorts design study [
      • Widom C.S.
      Childhood victimization early adversity, later psychopathology.
      ,
      • Widom C.S.
      The cycle of violence.
      ,
      • McGloin J.M.
      • Widom C.S.
      Resilience among abused and neglected children grown up.
      ]. Findings there are consistent with those from other studies; namely, individuals who were abused were found to be at significantly higher risk than controls for youth violent crime (arrests). Effects of abuse on violence continued into adulthood.

      Protection and resilience

      Evidence on the relation between physical abuse and youth antisocial behavior (delinquency and violence) is compelling. However, not all children who are maltreated subsequently engage in antisocial behavior; that is, they appear protected from this sequela of abuse [
      • McGloin J.M.
      • Widom C.S.
      Resilience among abused and neglected children grown up.
      ,
      • Herrenkohl E.C.
      • Herrenkohl R.C.
      • Egolf B.
      Resilient early school-age children from maltreating homes outcomes in late adolescence.
      ,
      • Masten A.S.
      Ordinary magic resilience processes in development.
      ]. The results from the Smith and Thornberry study above illustrate this point. In the study, 45% of maltreated youth engaged in subsequent delinquency; the majority of youth (55%), however, avoided later delinquency.
      Individuals who show better than expected outcomes having been exposed to some risk (such as abuse), or those able to avoid negative outcomes (such as antisocial behavior) in the face of adversity, often are described as “resilient” [
      • Masten A.S.
      Ordinary magic resilience processes in development.
      ]. Resilience in vulnerable children has links to individual characteristics, such as high IQ and positive temperament, and to social influences that modify, and in some cases ameliorate, the damage caused to children’s development by earlier risk exposure [
      • Herrenkohl E.C.
      • Herrenkohl R.C.
      • Egolf B.
      Resilient early school-age children from maltreating homes outcomes in late adolescence.
      ,
      • Masten A.S.
      Ordinary magic resilience processes in development.
      ,
      • Masten A.S.
      Resilience in individual development successful adaptation despite risk and adversity.
      ,
      • Rutter M.
      Psychosocial adversity risk, resilience and recovery.
      ,
      • Werner E.E.
      Resilience in development.
      ,
      • Masten A.S.
      • Best K.M.
      • Garmezy N.
      Resilience and development contributions from the study of children who overcome adversity.
      ]. For example, with regard to antisocial behavior as an outcome, Morrison et al. [
      • Morrison G.M.
      • Robertson L.
      • Laurie B.
      • Kelly J.
      Protective factors related to antisocial behavior trajectories.
      ] found that youth who experienced a high degree of social support, parent supervision, and classroom participation fared better than those who did not. The effects of these protective factors on behavior were maintained after accounting for child gender and earlier antisocial behavior, as well as other protective factors such as self-control and assertive problem-solving.
      Findings on resilience across a range of adversities do appear in the research literature. However, the quality of the research that has generated these findings is mixed. Many studies on resilience involve small samples or case studies, which limits the accuracy and generalizability of results [
      • Masten A.S.
      Ordinary magic resilience processes in development.
      ]. Also problematic is that researchers have studied resilience by examining protective influences only within a single, vulnerable group of children, failing to include a comparison group. From analyses, researchers have drawn conclusions about processes of resilience thought to be unique to high risk children when those processes could apply to other children, not just those who experienced a given form of adversity [
      • Masten A.S.
      Ordinary magic resilience processes in development.
      ].
      In the child maltreatment literature, more universal methodological shortcomings exist [
      • Tajima E.A.
      • Herrenkohl T.I.
      • Huang B.
      • Whitney S.D.
      Measuring child maltreatment a comparison of prospective parent reports and retrospective adolescent reports.
      ,
      • Henry B.
      • Moffitt T.E.
      • Caspi A.
      • Langley J.
      • Silva P.A.
      On the “remembrance of things past” a longitudinal evaluation of the retrospective method.
      ,
      • Widom C.S.
      Does violence beget violence? A critical examination of the literature.
      ,
      • Widom C.S.
      • Shepard R.L.
      Accuracy of adult recollections of childhood victimization: Part 1. Childhood physical abuse.
      ], which further undermine the strength of existing knowledge on resilience for abused children. Perhaps most troubling is the abundance of studies that use cross-sectional designs and retrospective measures of child maltreatment, absent validation against prospective measures [
      • Herrenkohl T.I.
      • Huang B.
      • Tajima E.A.
      • Whitney S.D.
      Examining the link between child abuse and youth violence an analysis of mediating mechanisms.
      ,
      • Tajima E.A.
      • Herrenkohl T.I.
      • Huang B.
      • Whitney S.D.
      Measuring child maltreatment a comparison of prospective parent reports and retrospective adolescent reports.
      ]. The use of cross-sectional designs leads to ambiguity in causal effects; retrospective measurement is problematic because assessment depends on accurate recall of distant childhood events, which are subject to distortion and/or selectively recalled. Sole reliance on official records to measure abuse, though providing prospective measurement, dramatically underestimates maltreatment cases, which again affects generalizability [
      • Straus M.A.
      • Smith C.
      Family patterns and child abuse.
      ]. The longitudinal design of the Lehigh Longitudinal Study and its use of multiple sources of data on child maltreatment (official records, reports from caregivers about their punitive, abusive disciplining of their children, and youth retrospective reports) offer here a solid base from which to study the consequences of child maltreatment and resilience in children who have been physically abused.

      Objectives and rationale

      The current study examines several factors hypothesized to protect physically maltreated children from later antisocial behavior—status offenses, violence, and delinquency. This study incorporates factors associated with an individual child (positive future orientation), influence from parents and peers (parent and peer disapproval of antisocial behavior), and a child’s involvement in, and commitment to school and religion. Several of these or related factors (e.g., bonding to school and religiosity) are hypothesized elsewhere to affect risk for youth delinquency and violence and are embedded within social developmental theories on these topics [
      • Catalano R.F.
      • Hawkins J.D.
      The social development model a theory of antisocial behavior.
      ,
      • Hawkins J.D.
      • Herrenkohl T.
      • Farrington D.P.
      • et al.
      A review of predictors of youth violence.
      ,
      • Herrenkohl T.I.
      • Hill K.G.
      • Chung I.-J.
      • et al.
      Protective factors against serious violent behavior in adolescence a prospective study of aggressive children.
      ]. The protective factors were chosen for the study to represent an ecological framework (e.g., Bronfenbrenner [
      • Bronfenbrenner U.
      ]) that situates an individual child within social contexts; such a perspective hypothesizes that developmental outcomes for maltreated children emerge as much from a child’s environmental surroundings as they do from his/her individual qualities. Few studies have investigated such factors in relationship to the risk posed by child maltreatment specifically. Moreover, given the previously noted methodological weaknesses of existing studies, further investigation of these and other protective factors is warranted.
      So as to examine the degree to which protective influences are unique to maltreated children, or shared among other children in the sample (that is, those not meeting criteria for abuse), we include analyses for comparison children. We also investigate the salience of protective factors for a smaller subgroup of abused children who are at risk for antisocial behavior on other factors in addition, namely low family socioeconomic status and earlier (childhood) antisocial behavior. This additional step allows for examination of protective influences within a highly vulnerable group of children. Collectively, analyses should provide a basis for judging the degree to which protective influences are salient for maltreated children alone or are generalizable to children at higher and lower levels of risk for antisocial behavior.

      Methods

      Sample

      Study participants were sampled from child welfare abuse and protective service programs, Head Start centers, and from child care programs in a two-county area of Pennsylvania; this two-county area included a substantial urban/suburban population as well as some rural areas. Selection of the sample of abusive and neglectful child welfare families was accomplished by county child welfare agencies referring to the study, over a 2-year period, all new and some ongoing cases in which there was at least one abused or neglected child 18 months to 6 years of age in the home. Families were informed of the study by the agency and were approached by a member of the project staff to request their participation. Children in the two child welfare groups participated in one of several group settings (e.g., day care family and classroom settings and Head Start classrooms). It was from these settings that child participants who comprise the other groups in the sample were obtained. These settings were geographically spread over the same two counties served by the child welfare agencies referring to the study and included 13 Head Start centers, 12 day care programs, two programs for handicapped children, three Home Start programs, and eight nursery school programs.
      An initial assessment of children and their families was completed in 1976–1977, when children were of preschool age. Children then ranged in age from 18 months to 6 years. A second assessment was completed in 1980–1982, when the children were in elementary school. Eighty-two percent of children assessed in preschool were again assessed in elementary school. A third and final assessment was completed in 1990–1992, when children were adolescents (average age = 18 years). That assessment included 416 (91%) of the original 457 children. Parents consented in writing in the first two waves of the study and were informed about confidentiality. Consent from adolescent children was obtained in writing in the third wave of the study.
      The full sample (N = 457) contains 248 (54%) males and 209 females from 297 families. One child was assessed in 52% (n = 155) of the families; two children were assessed in 43% (n = 128) of the families; three or four children were assessed in 5% (n = 14) of the families. Prior analyses that tested the effects of case clustering within families on the estimation of parameters found no substantive differences in findings when variability within families was/was not directly modeled, likely owing to the small number of families (5%) with more than two children present in the sample. The race breakdown of the full sample is: 1.3% (n = 6) American Indian/Alaska Native, .2% (n = 1) Native Hawaiian or Other Pacific Islander, 5.3% (n = 24) black or African-American, 80.7% (n = 369) white, 11.2% (n = 51) more than one race, and 1.3% (n = 6) unknown. The ethnic composition is: 7.1% (n = 33) Hispanic or Latino, 91.5% (n = 381) Not Hispanic or Latino, and 1.3% (n = 6) unknown. These percentages are consistent with the makeup of the two-county area from which the sample was drawn. Eighty-six percent of children were, at the time of initial assessment, from two-parent households. Sixty-three percent of families had incomes below $700 per month in 1976–1977.
      Of the 416 participants assessed in adolescence, 229 (55.0%) are males, 1.4% (n = 6) are American Indian/Alaska Native, .2% (n = 1) Native Hawaiian or Other Pacific Islander, 5.0% (n = 21) black or African-American, 81.5% (n = 339) white, and 11.7% (n = 49) more than one race. By the time of this assessment, four participants had died: one child in the child welfare abuse group had died from illness and another from a car accident; one in the child welfare neglect group had been murdered; and one child in the middle-income group had died in a car accident.
      The equality of attrition across groups (in adolescence) was tested. Percentages lost to attrition from each composite group in the sample (e.g., child welfare abuse) were not significantly different. Further, no significant differences were found when those lost to attrition and those who remained in the study were compared on several key variables, including childhood SES and severity of physical discipline.
      Data for the preschool- and school-age assessments are from interviews with parents. Interviews produced data on a range of family and child variables, including information about parents’ disciplining practices. In the assessment, parents reported on their use of 39 discipline practices that ranged in severity from verbal reprimands and spanking to burning a child and physical assault resulting in serious injury.
      Data for the adolescent assessment are from face-to-face interviews and individually administered questionnaires with youth and parents. The youth survey provides information on parenting practices, youth behavior, psychological functioning, school achievement, studying habits, educational aspirations, suspensions and disciplinary action, and school dropout. Data on social skills, peer group interactions, and family relations also are available. Interviews with parents provide additional information on many of these same constructs.

      Measures

      Child age, gender (1 = male, 2 = female), and early antisocial behavior were included as controls in analyses with the abuse group and the “no-abuse” comparison group. Early antisocial behavior and a measure of family socioeconomic status (SES) were used later to define the multi-risk subgroup drawn from within the sample of maltreated children. SES is a composite variable that defines mother’s occupational status and education level, family income in the preschool period, and total rooms in a family’s house. Each indicator on this scale was standardized before all were summed, providing a single composite SES score for each child.
      Our measure of early antisocial behavior was derived from mothers’ reports of their children’s conduct problems at the time of the school-age assessment. Using a modified version of the Achenbach Child Behavior Checklist (CBC) [
      • Achenbach T.M.
      The Child Behavior Profile: I. Boys aged 6–11.
      ,
      • Achenbach T.M.
      • Edelbrock C.S.
      The Child Behavior Profile: II. Boys aged 12–16 and girls aged 6–11 and 12–16.
      ] (to the scale, several positive behavior indicators were added to complement the otherwise negative behavior items that appear on the CBC), mothers reported on their child’s aggressive (18 items: e.g., teases, cruel or mean to others, destroys things) and delinquent (10 items: e.g., vandalizes, steals, runs away) behaviors. Items for the two subscales (α = .84 and .71, respectively) were standardized and added to form a single measure.
      Physical abuse was assessed using three data sources: (a) official records of substantiated abuse cases, (b) mothers’ reports of their disciplining of their preschool- and school-age children, and (c) adolescents’ retrospective reports of those same discipline practices. Behaviors assessed with self-reports from mothers and adolescent children were: biting a child; slapping so as to bruise a child; hitting a child with a stick, paddle or other hard object; or hitting a child with a strap, rope, or belt. In each case, the number of practices to which a child was exposed was documented. Those who were disciplined with two or more abusive practices were considered to have been maltreated according to that (self-report) measure. A threshold of two or more incidents was set to eliminate isolated cases of severe physical disciplining from an otherwise nonabusive parent. Individuals for whom there was agreement between the prospective and retrospective indicators of abuse were added to those identified as abuse victims through official records. This procedure allowed us to take advantage of the multiple sources of data available in the study. By requiring evidence of abuse on both the prospective and retrospective measures before identifying a child as a victim of abuse on these self-report indicators, we lessen the potential error in measurement that might be introduced by one or the other source [
      • Tajima E.A.
      • Herrenkohl T.I.
      • Huang B.
      • Whitney S.D.
      Measuring child maltreatment a comparison of prospective parent reports and retrospective adolescent reports.
      ]. With the combined data, analyses arrived at a group of abused (n = 176, 42.3% of the sample of 416) and nonabused (n = 240, 57.7% of the sample of 416) children. Demographics for the two groups are shown in Table 1.
      Table 1Descriptive information for the comparison, abuse, and multi-risk groups
      VariablesComparison (n = 240)Abuse (n = 176)Multi-risk
      The multi-risk group includes a subgroup of cases from the larger “abuse” group.
      (n = 58)
      Gender
       Male131 (54.6%)98 (55.7%)36 (62.1%)
       Female109 (45.4%)78 (44.3%)22 (37.9%)
      Age
       Mean17.8518.7618.59
       (SD)(1.77)(1.74)(1.70)
      Race
       White198 (82.5%)137 (77.8%)45 (77.6%)
       Black13 (5.4%)7 (4.0%)2 (3.4%)
       Multiracial21 (8.8%)28 (15.9%)9 (15.5%)
       Other/Unknown8 (3.4%)4 (2.3%)2 (3.4)
      SES
      SES is a composite variable that defines mother’s occupational status and education level, family income in the preschool period, and total rooms in a family’s house. Each indicator on this scale was standardized before all were summed, providing a single composite SES score for each child.
       Mean.24−.28−.49
       (SD)(.95)(.48)(.29)
      a The multi-risk group includes a subgroup of cases from the larger “abuse” group.
      b SES is a composite variable that defines mother’s occupational status and education level, family income in the preschool period, and total rooms in a family’s house. Each indicator on this scale was standardized before all were summed, providing a single composite SES score for each child.
      A final set of analyses examined protective factors within a “multi-risk” subgroup (n = 58) drawn from the analysis sample of 176 abused children. This multi-risk group scored below the sample mean on SES and above the mean for the full sample on early antisocial behavior. Demographics for this subgroup also are shown in Table 1.
      In these later analyses, early antisocial behavior was removed as a control variable; however, gender and age of the child continued to be used as controls.

      Adolescent antisocial behavior

      Analyses included three measures of antisocial behavior: Violence is a count of four lifetime acts, which includes reports of having been involved in gang fighting, having hit someone other than siblings or a parent, having had intent to seriously hurt or kill another person, and having forced someone to have sex against their will (α = .57). This measure has a mean of 1.24 (SD = 1.01) and a range of 0–4 for the sample. Delinquency combined data on youths’ reports of violence with other law-violating behaviors, such as stealing, breaking and entering, and damaging property. The measure included a total of 34 items (α = .92) and has a mean of 10.84 (SD = 7.73) and a range of 0–34 for the sample. Status offenses is a count of seven lifetime acts. Youths indicated whether they had run away from home, skipped school, been suspended or expelled, drunk alcohol, driven a car without a license, or been reported to juvenile authorities (α = .64). This measure has a mean of 3.24 (SD = 1.64) and a range of 0–7 for the sample. Items for the violence and delinquency scales are from the Elliott et al. [
      • Elliott D.S.
      • Dunford F.W.
      • Huizinga D.
      The identification and prediction of career offenders utilizing self-reported and official data.
      ] self-reported delinquency scale, originally developed for the National Youth Survey (NYS); the scales in the current study correspond with those used in that original study. The status offenses measure was developed for the current study to capture less serious forms of antisocial behavior that would still carry notable consequences for youth violators. Although these measures are positively correlated (Table 2), there are conceptually meaningful differences among them, especially with regard to the nature and seriousness of the behaviors they represent [
      • Hawkins J.D.
      • Herrenkohl T.
      • Farrington D.P.
      • et al.
      A review of predictors of youth violence.
      ,
      • Elliott D.S.
      • Dunford F.W.
      • Huizinga D.
      The identification and prediction of career offenders utilizing self-reported and official data.
      ].
      Table 2Correlations among the variables
      12345678910111213
      SES−.34
      p < .001.
      −.01−.28
      p < .001.
      −.31
      p < .001.
      .33
      p < .001.
      .08
      p < .10;
      .33
      p < .001.
      .01.30
      p < .001.
      −.29
      p < .001.
      −.20
      p < .001.
      −.30
      p < .001.
      Age.00.14
      p < .01;
      .25
      p < .001.
      −.14
      p < .01;
      −.09
      p < .10;
      −.26
      p < .001.
      .02−.20
      p < .001.
      .21
      p < .001.
      .29
      p < .001.
      .29
      p < .001.
      Gender−.24
      p < .001.
      −.01.12
      p < .05;
      .15
      p < .01;
      .04.12
      p < .05;
      .17
      p < .01;
      −.32
      p < .001.
      −.36
      p < .001.
      −.23
      p < .001.
      Early behavior.18
      p < .01;
      −.17
      p < .01;
      −.09−.18
      p < .01;
      −.05−.19
      p < .001.
      .33
      p < .001.
      .32
      p < .001.
      .33
      p < .001.
      Abuse−.10
      p < .05;
      −.10
      p < .10;
      −.19
      p < .001.
      −.05−.19
      p < .001.
      .33
      p < .001.
      .32
      p < .001.
      .33
      p < .001.
      School commitment.27
      p < .001.
      .32
      p < .001.
      .22
      p < .001.
      .70
      p < .001.
      −.28
      p < .001.
      −.29
      p < .001.
      −.33
      p < .001.
      Parent/peer disapproval.27
      p < .001.
      .16
      p < .01;
      .67
      p < .001.
      −.24
      p < .001.
      −.34
      p < .001.
      −.30
      p < .001.
      Religion.06.64
      p < .001.
      −.24
      p < .001.
      −.25
      p < .001.
      −.26
      p < .001.
      Positive future orientation.56
      p < .001.
      .12
      p < .05;
      −.14
      p < .01;
      −.11
      p < .05;
      Protective count−.36
      p < .001.
      −.41
      p < .001.
      −.40
      p < .001.
      Violence.72
      p < .001.
      .59
      p < .001.
      Delinquency.70
      p < .001.
      Status offenses
      + p < .10;
      low asterisk p < .05;
      low asterisklow asterisk p < .01;
      low asterisklow asterisklow asterisk p < .001.

      Protective factors

      We hypothesized in this study that risk for antisocial behavior would be lessened by youths’ commitment to school and importance of their education (e.g., youth are satisfied with and value their education; spend time studying; 11 items, α = .84); parent/peer disapproval of antisocial behavior (e.g., parents and/or peers communicate that it is not OK to use alcohol or drugs, steal, or perpetrate violence; 20 items, α = .89); an individual’s positive future orientation, such as having expectations to succeed, be happy, and have good times in the future, (16 items, α = .70); and participation in, and importance of, religion (e.g., youth attend a religious institution and view religion as important; 2 items, r = .53). Each variable was dichotomized to isolate scores in the top 25% of each scale distribution, or as near to that point as possible. This procedure simplifies interpretation of analysis results (by forming a binary measure reflecting presence or absence of a given factor) without an appreciable loss of information [
      • Farrington D.P.
      • Loeber R.
      Some benefits of dichotomization in psychiatric and criminological research.
      ,
      • Loeber R.
      • Farrington D.P.
      • Stouthamer-Loeber M.
      • Van Kammen W.B.
      Multiple risk factors for multiproblem boys co-occurrence of delinquency, substance use, attention deficit, conduct problems, physical aggression, covert behavior, depressed mood, and shy/withdrawn behavior.
      ]. Further, there is evidence that predictors of antisocial behavior examined in their dichotomous and continuous forms yield similar results [
      • Herrenkohl T.I.
      • Maguin E.
      • Hill K.G.
      • et al.
      Developmental risk factors for youth violence.
      ]. Indeed, when analyses (not reported) compared the effects of dichotomous and continuous measures in the current study, similar results were achieved.
      A correlation of all measures used in the study, including the variable that defined the abuse and comparison groups (0 = no abuse, 1 = abuse), protective factors, and the three behavior outcomes are shown in Table 2. A full listing of items that comprise each measure is available upon request from the corresponding author.

      Analysis

      All analyses used a two-step hierarchical regression approach. In step 1, age, gender, and early antisocial behavior (omitted in the later multi-risk subgroup analyses) were entered as controls. Analyses that involved the multi-risk subgroup within the analysis sample of abused children included only child gender and age as controls; early antisocial behavior was, in this case, used to define the group. In step 2, each protective factor (school commitment/importance, parent/peer disapproval of antisocial behavior, positive future orientation, and religion) was entered separately as a predictor; its conditional effect (adjusted for controls in the model) on each outcome was observed. A significant negative association between the predictor and outcome (and significant change in the amount of variance explained) was taken to mean a lessening of risk associated with exposure to that (protective) factor. A final regression model in each case examined the additive (combined) negative effect of all protective factors on a given outcome; this analysis was done to ascertain whether the presence of more protective factors resulted in a stronger (inverse) association with antisocial behavior. All analyses were run for youth in the “abuse” group and then again for youth in the “no-abuse” comparison group. Tests of predictor-by-group interactions were used to determine whether protective effects are unique to those who had been abused, or generalizable to comparison youth.
      A later set of regression models tested protective factors for a multi-risk subgroup of abused children (abused, low SES, and high early antisocial behavior). Both low SES and early antisocial behavior were used to define this multi-risk group because they appear often in the research literature as salient predictors of antisocial behavior [
      • Hawkins J.D.
      • Herrenkohl T.
      • Farrington D.P.
      • et al.
      A review of predictors of youth violence.
      ,
      • Herrenkohl T.I.
      • Hill K.G.
      • Chung I.-J.
      • et al.
      Protective factors against serious violent behavior in adolescence a prospective study of aggressive children.
      ]. Again, to assess for invariance in protective effects across the (abuse and multi-risk) groups, analyses examined protective factor-by-group interactions. In the analysis, multi-risk children were compared with those at-risk on abuse alone. Absence of significant interactions would argue for combining the groups in tests of protective factor influences.

      Results

      Examination of protective factors within the abuse and comparison groups

      Table 3 contains the results of all analyses across the three outcome variables—violence, delinquency, and status offenses—for the abuse group and the comparison (no-abuse) group. For each behavior, the effect of each protective factor variable (standardized and unstandardized regression coefficients) for each group (abuse group is the top number in each cell) is shown. Also included in the table is a listing of the change in total variance explained (ΔR2) with the addition of each protective factor after accounting for controls.
      Table 3Relations between protective factors and youth outcomes for the abuse and comparison groups
      Protective factorGroupViolenceDelinquencyStatus offenses
      St. BB(SE)[Δ R2]St. BB(SE)[Δ R2]St. BB(SE)[Δ R2]
      School commitment and importanceAbuse Comparison−.18 −.13−.45 −.28(.20)
      p < .05;
      (.14)
      p < .05;
      .03 .02−.22 −.14−4.08 −2.34(1.43)
      p < .01;
      (1.07)
      p < .05;
      .05 .02−.26 −.19−1.02 −.64(.29)
      p < .01;
      (.23)
      p < .01;
      .07 .03
      Parent and peer disapproval of antisocial behavioraAbuse Comparison−.10 −.16−.25 −.36(.20) (.14)
      p < .05;
      .01 .03−.29 −.26−5.73 −4.24(1.43)
      p < .001.
      (1.02)
      p < .001.
      .08 .06−.13 −.27−.51 −.90(.31) (.22)
      p < .001.
      .02 .07
      Positive future orientationAbuse Comparison−.12 −.08−.31 −.17(.20) (.13).02 .01−.18 −.04−3.31 −.66(1.47)
      p < .05;
      (1.01)
      .03 .002−.10 −.02−.40 −.08(.31) (.22).01 .001
      ReligionAbuse−.16−.43(.23)
      p < .10;
      .02−.03−.63(1.72).001−.08−.35(.35).01
      Comparison−.12−.26(.15)
      p < .10;
      .01−.21−3.51(1.08)
      p < .01;
      .04−.20−.68(.23)
      p < .01;
      .04
      Protective factor countAbuse Comparison−.25 −.21−.27 −.17(.09)
      p < .01;
      (.06)
      p < .01;
      .06 .07−.31 −.26−2.58 −1.64(.63)
      p < .001.
      (.41)
      p < .001.
      .09.06−.24 .30−.41 −.38(.13)
      p < .01;
      (.09)
      p < .001.
      .06 .08
      St. B is the standardized regression coefficient; B is the unstandardized coefficient.
      + p < .10;
      low asterisk p < .05;
      low asterisklow asterisk p < .01;
      low asterisklow asterisklow asterisk p < .001.
      As shown, for both the abuse and comparison groups, significant main effects (p < .05) of several protective factors emerged. For example, school commitment and importance predicted less of each outcome: violence, delinquency, and status offenses. Parent and peer disapproval of antisocial behavior also was highly (inversely) predictive of delinquency in both groups. In contrast, positive future orientation failed to predict violence or status offenses, although it was related to delinquency for abused children. For both abused and comparison children, the overall number of protective factors was significantly associated with each outcome; that is, the more protective factors (0–4) to which a child was exposed during adolescence, the lower was his/her involvement in each behavior examined. Depending on the outcome, this variable explained an additional 4% to 9% of variance after controlling for age, gender, and early antisocial behavior.
      When tested, no protective factor-by-group interactions were found, suggesting that, although there appear to be some differences in the magnitude of protective influences for the two groups, the differences are not large enough to produce significant findings.

      Examination of protective factors within the multi-risk group

      The next step was to examine the same protective factors for the subgroup of children drawn from within the abuse sample to represent a multi-risk category. These children were abused and were below the sample mean on family SES; they also were above the sample mean on early antisocial behavior. Here, we examined interactions between each protective factor and a variable specifying the “multi-risk” and “abuse” subgroups. As before, we found no significant interactions between the groups, suggesting that the strength of protective influences is, in essence, the same for both groups.
      In light of the above findings, we re-estimated the effects of all protective factors for the larger youth sample, combining the abuse (including multi-risk children) and comparison subgroups. The results, shown in Table 4, are thus “overall” effects of the protective factors in relationship to violence, delinquency, and status offenses.
      Table 4Relations between protective factors and youth outcomes for the combined sample
      Protective factorViolenceDelinquencyStatus Offenses
      St. BB(SE)[Δ R2]St. BB(SE)[Δ R2]St. BB(SE)[Δ R2]
      School commitment and importance−.17−.39(.11)
      p < .01;
      .03−.18−3.21(.85)
      p < .001.
      .03−.22−.81(.18)
      p < .001.
      .05
      Parent and peer disapproval of antisocial behavior−.15−.35(.12)
      p < .01;
      .02−.27−4.89(.83)
      p < .001.
      .07−.21−.78(.18)
      p < .001.
      .04
      Positive future orientationa−.10−.22(.11)
      p < .05;
      .01−.10−1.71(.84)
      p < .05;
      .01−.07−.23(.18)
      p < .05;
      .004
      Religion−.15−.35(.12)
      p < .01;
      .02−.15−2.68(.92)
      p < .01;
      .02−.16−.59(.19)
      p < .01;
      .02
      Protective factor count−.24−.22(.05)
      p < .001.
      .05−.29−2.03(.34)
      p < .001.
      .08−.28−.39(.07)
      p < .001.
      .07
      low asterisk p < .05;
      low asterisklow asterisk p < .01;
      low asterisklow asterisklow asterisk p < .001.
      Except for the effect of positive future orientation on status offenses, all protective factors were significantly related to the outcomes in the hypothesized direct. Particularly robust effects, as evidenced in part by the proportionally large percentage of variance in the outcome explained by the variable, are shown for school commitment and importance and parent and peer disapproval of antisocial behavior. Positive future orientation and religion, although mostly significant, were less strongly predictive of antisocial behavior. Again, the overall number of protective factors was, however, very strongly predictive of violence, delinquency, and status offenses each. Here, the variable explained from 5% to 8% of additional variance in an outcome after accounting for age and gender.

      Discussion

      Results of this study show that among both abused and non-abused children, having a strong commitment to school (and viewing one’s education as important), having parents and peers who disapprove of antisocial behavior, and being involved in a religious community may each independently lower risk for antisocial behavior during adolescence. Exposure to an increasing number of these protective factors resulted in a notable lowering of risk for antisocial behavior among children after accounting for demographic controls (age, gender) and early antisocial behavior.
      Findings of invariance across groups in this study are important because they work against a commonly held belief that children who overcome extreme adversity possess qualities that set them apart from other children [
      • Masten A.S.
      Ordinary magic resilience processes in development.
      ]. Instead, this study suggests that children who are abused (including those exposed to other risk factors), and those not, respond similarly to developmental experiences that work against antisocial behavior. Thus, what likely differentiates at-risk children who achieve resilience from those who do not is their level of exposure to positive, yet normative, developmental factors. Unfortunately, research appears to suggest that children at highest risk for antisocial behavior are precisely those children who experience the lowest levels of protection because they are socialized in resource-deficient environments [
      • Herrenkohl T.I.
      • Hill K.G.
      • Chung I.-J.
      • et al.
      Protective factors against serious violent behavior in adolescence a prospective study of aggressive children.
      ,
      • Pollard J.A.
      • Hawkins J.D.
      • Arthur M.W.
      Risk and protection are both necessary to understand diverse behavioral outcomes in adolescence?.
      ]. Thus, intervening with in these environments directly to enhance protection is extremely important.
      Findings provide further evidence that the degree to which children are at risk for antisocial behavior during adolescence depends, in part, on the extent to which they invest in their schooling and become bonded to the institution of school. Although in Morrison et al’s [
      • Morrison G.M.
      • Robertson L.
      • Laurie B.
      • Kelly J.
      Protective factors related to antisocial behavior trajectories.
      ] study on protection from antisocial behavior referenced earlier, school bonding was not shown to have a unique effect on the outcome after controlling for gender, earlier problem behavior, and several other hypothesized factors, results elsewhere are in strong contrast [
      • Catalano R.F.
      • Hawkins J.D.
      The social development model a theory of antisocial behavior.
      ,
      • Herrenkohl T.I.
      • Hill K.G.
      • Chung I.-J.
      • et al.
      Protective factors against serious violent behavior in adolescence a prospective study of aggressive children.
      ,
      • Hawkins J.D.
      • Herrenkohl T.I.
      Prevention in the school years.
      ]. For example, in a study by Herrenkohl et al. [
      • Herrenkohl T.I.
      • Hill K.G.
      • Chung I.-J.
      • et al.
      Protective factors against serious violent behavior in adolescence a prospective study of aggressive children.
      ] from the Seattle Social Development Project, children with early aggressive behavior were protected from later violence by strong school bonds. In fact, the probability of violence among aggressive children who later became bonded to school, and those who did not, differed by 20%; youth bonded to school had a 20% probability of serious violence at age 18, those not bonded to school had a 40% probability of later violence. For prevention, it thus appears that by strengthening children’s connection to school, potential for negative youth outcomes (such as antisocial behavior) may be lessened [
      • Hawkins J.D.
      • Herrenkohl T.I.
      Prevention in the school years.
      ,
      • Herrenkohl T.I.
      • Chung I.-J.
      • Catalano R.F.
      Review of research on predictors of youth violence and school-based and community-based prevention approaches.
      ].
      In this study, having parents and peers who disapproved of youths’ involvement in antisocial behavior also appeared protective. Thus, efforts to engage parents and peers in the process of discouraging antisocial behavior might have beneficial effects for vulnerable children. It is possible that this process would have less a direct or immediate effect on youth in families where parent-child bonds are strained by a history of abuse, however, research has yet to establish this to be the case.
      Being involved in a religious institution and perceiving religion as important also appears to lessen risk for violence, delinquency, and status offenses for abused and non-abused children, if less strongly than the previous two factors (school commitment and parent/peer disapproval). Perhaps most beneficial to youth in this regard are the opportunities religious settings provide for youth to give and receive social support from others within an easily accessible community setting. Also potentially important are the positive messages children and youth receive within such settings regarding tolerance and peaceful resolution of problems.
      Having a positive outlook on the future was less protective against antisocial behavior in this study. This finding appears consistent with other research that has documented less robust protective effects of individual/constitutional factors, including self-esteem and self-concept (see one example in Morrison et al. [
      • Morrison G.M.
      • Robertson L.
      • Laurie B.
      • Kelly J.
      Protective factors related to antisocial behavior trajectories.
      ]). Thus, research appears to suggest that although efforts should continue to seek individual sources of resilience among vulnerable children, as much or more attention should be focused on social and contextual influences. Given that other individual variables shown to reliably predict problem behavior in youth, such as IQ and temperament, are less malleable through preventive intervention, a focus on social and contextual sources of resilience and protection is all the more important.
      In this study, exposure to an increasing number of protective factors resulted in lower levels of antisocial behavior for youth in subgroup and overall tests. This finding coincides with those from earlier studies on protection and violence [
      • Herrenkohl T.I.
      • Hill K.G.
      • Chung I.-J.
      • et al.
      Protective factors against serious violent behavior in adolescence a prospective study of aggressive children.
      ,
      • Pollard J.A.
      • Hawkins J.D.
      • Arthur M.W.
      Risk and protection are both necessary to understand diverse behavioral outcomes in adolescence?.
      ]. Thus, it follows that comprehensive prevention efforts (those that target multiple protective [and risk] factors) likely will benefit youth more than will efforts that seek to address a single, stand-alone predictor [
      • Herrenkohl T.I.
      • Chung I.-J.
      • Catalano R.F.
      Review of research on predictors of youth violence and school-based and community-based prevention approaches.
      ,
      • Fraser M.W.
      • Richman J.M.
      Resilience implications for evidence-based practice.
      ]. Indeed, such a multifaceted intervention approach has demonstrated effects in lessening problem behavior of all sorts, including those examined in this study [
      • Hawkins J.D.
      • Herrenkohl T.I.
      Prevention in the school years.
      ,
      • Herrenkohl T.I.
      • Chung I.-J.
      • Catalano R.F.
      Review of research on predictors of youth violence and school-based and community-based prevention approaches.
      ,
      • Fraser M.W.
      • Richman J.M.
      Resilience implications for evidence-based practice.
      ,
      • Wasserman G.A.
      • Miller L.S.
      The prevention of serious and violent juvenile offending.
      ].

      Limitations

      Perhaps the most notable limitation of this study is that protective factors were examined in separate models, although a final model did in each case examine a combination of all protective factors. Although analyses did control for demographics and earlier antisocial behavior among children, they did not investigate the interactions among protective variables as they affect youth outcomes, nor did they investigate overlap in the outcomes themselves. Further investigation of these and other protective variables in multivariate analyses is warranted. Also warranted are further investigations of outcomes other than those examined here that can be used to characterize child resilience. A lessening of risk for antisocial behavior is but one indicator of positive youth functioning. Analyses of data on outcomes that pertain to broader dimensions of the physical and emotional well-being of youth are very much needed.
      Other potential limitations of this study are reliance on self-report data in assessing antisocial behavior and contemporaneous measurement of protective factors and outcomes. Although use of contemporaneous measures allows for examination of proximal influences on behavior, they limit the degree to which causal inferences about predictors and outcomes can be made. Finally, the sample used in this study is primarily white. Thus, results of this study may not generalize to other races and ethnic groups.

      Conclusions

      The findings in this study are important because they pertain to protective influences for abused, multi-risk, and comparison children derived from data that are methodologically sound. Findings also are important because they provide implications for prevention. The protective factors that were investigated in this study represent targets for preventive intervention that are viable for children as they enter adolescence, a developmental period characterized by transition and increasing experimentation with a full range of behaviors, both prosocial and antisocial. The fact that protective factors were predictive of lower antisocial behavior in both the abuse and comparison groups suggests that protective effects are more universal than they are unique to a given group of children. Thus, universal interventions that seek to lessen risk for negative outcomes by enhancing protection among all children appear justified.

      Acknowledgments

      Work on this project was supported by funds from the Social Work Prevention Research Center, School of Social Work, University of Washington (National Institute of Mental Health grant R24MH56599, Lewayne Gilchrist, PI) and by the University of Washington Royalty Research Fund.

      References

        • Widom C.S.
        Childhood victimization.
        Natl Inst Justice J. 2000; 242: 2-9
        • Fergusson D.M.
        • Lynskey M.T.
        Physical punishment/maltreatment during childhood and adjustment in young adulthood.
        Child Abuse Negl. 1997; 21: 617-630
        • Herrenkohl R.C.
        • Egolf B.P.
        • Herrenkohl E.C.
        Preschool antecedents of adolescent assaultive behavior.
        Am J Orthopsychiatry. 1997; 67: 422-432
        • Herrenkohl T.I.
        • Huang B.
        • Tajima E.A.
        • Whitney S.D.
        Examining the link between child abuse and youth violence.
        J Interpers Violence. 2003; 18: 1189-1208
        • Smith C.
        • Thornberry T.P.
        The relationship between childhood maltreatment and adolescent involvement in delinquency.
        Criminology. 1995; 33: 451-481
        • Stouthamer-Loeber M.
        • Loeber R.
        • Homish D.
        • Wei E.
        Maltreatment of boys and the development of disruptive and delinquent behavior.
        Dev Psychopathol. 2001; 13: 941-955
        • Tajima E.A.
        • Herrenkohl T.I.
        • Huang B.
        • Whitney S.D.
        Measuring child maltreatment.
        Am J Orthopsychiatry. 2004; 74: 424-435
        • Widom C.S.
        The cycle of violence.
        Science. 1989; 244: 160-166
        • McGloin J.M.
        • Widom C.S.
        Resilience among abused and neglected children grown up.
        Dev Psychopathol. 2001; 13: 1021-1038
        • Herrenkohl E.C.
        • Herrenkohl R.C.
        • Egolf B.
        Resilient early school-age children from maltreating homes.
        Am J Orthopsychiatry. 1994; 64: 301-309
        • Masten A.S.
        Ordinary magic.
        Am Psychol. 2001; 56: 227-238
        • Masten A.S.
        Resilience in individual development.
        in: Wang M.C. Gordon E.W. Educational Resilience in Inner City America Challenges and Prospects. Lawrence Erlbaum Associates, Inc, Hillsdale, NJ1994: 3-25
        • Rutter M.
        Psychosocial adversity.
        in: Richman J.M. Fraser M.W. The Context of Youth Violence Resilience, Risk and Protection. Praeger Publishers, Westport, CT2001: 13-42
        • Werner E.E.
        Resilience in development.
        Curr Direct Psychol Sci. 1995; 4: 81-85
        • Masten A.S.
        • Best K.M.
        • Garmezy N.
        Resilience and development.
        Dev Psychopathol. 1990; 2: 425-444
        • Morrison G.M.
        • Robertson L.
        • Laurie B.
        • Kelly J.
        Protective factors related to antisocial behavior trajectories.
        J Clin Psychol. 2002; 58: 277-290
        • Henry B.
        • Moffitt T.E.
        • Caspi A.
        • Langley J.
        • Silva P.A.
        On the “remembrance of things past”.
        Psychol Assess. 1994; 6: 92-101
        • Widom C.S.
        Does violence beget violence? A critical examination of the literature.
        Psychol Bull. 1989; 106: 3-28
        • Widom C.S.
        • Shepard R.L.
        Accuracy of adult recollections of childhood victimization: Part 1. Childhood physical abuse.
        Psychol Assess. 1996; 8: 412-421
        • Straus M.A.
        • Smith C.
        Family patterns and child abuse.
        in: Straus M. Gelles R. Physical Violence in American Families Risk Factors and Adaptations to Violence in 8,145 Families. Transaction Publishers, New Brunswick, NJ1990: 403-424
        • Catalano R.F.
        • Hawkins J.D.
        The social development model.
        in: Hawkins J.D. Delinquency and Crime Current Theories. Cambridge University Press, New York, NY1996: 149-197
        • Hawkins J.D.
        • Herrenkohl T.
        • Farrington D.P.
        • et al.
        A review of predictors of youth violence.
        in: Loeber R. Farrington D.P. Serious and Violent Juvenile Offenders Risk Factors and Successful Interventions. Sage, Thousand Oaks, CA1998: 106-146
        • Herrenkohl T.I.
        • Hill K.G.
        • Chung I.-J.
        • et al.
        Protective factors against serious violent behavior in adolescence.
        Soc Work Res. 2003; 27: 179-191
        • Bronfenbrenner U.
        The Ecology of Human Development. Harvard University Press, Cambridge, MA1979
        • Achenbach T.M.
        The Child Behavior Profile: I. Boys aged 6–11.
        J Consult Clin Psychol. 1978; 46: 478-488
        • Achenbach T.M.
        • Edelbrock C.S.
        The Child Behavior Profile: II. Boys aged 12–16 and girls aged 6–11 and 12–16.
        J Consult Clin Psychol. 1979; 47: 223-233
        • Elliott D.S.
        • Dunford F.W.
        • Huizinga D.
        The identification and prediction of career offenders utilizing self-reported and official data.
        in: Burchard J.D. Burchard S.N. Prevention of Delinquent Behavior. Vermont Conference on the Primary Prevention of Psychopathology. Vol. 10. Sage, Thousand Oaks, CA1987: 90-121
        • Farrington D.P.
        • Loeber R.
        Some benefits of dichotomization in psychiatric and criminological research.
        Crim Behav Ment Health. 2000; 10: 100-122
        • Loeber R.
        • Farrington D.P.
        • Stouthamer-Loeber M.
        • Van Kammen W.B.
        Multiple risk factors for multiproblem boys.
        in: Jessor R. New Perspectives on Adolescent Risk Behavior. Cambridge University Press, New York, NY1998: 90-149
        • Herrenkohl T.I.
        • Maguin E.
        • Hill K.G.
        • et al.
        Developmental risk factors for youth violence.
        J Adolesc Health. 2000; 26: 176-186
        • Pollard J.A.
        • Hawkins J.D.
        • Arthur M.W.
        Risk and protection.
        Soc Work Res. 1999; 23: 145-158
        • Hawkins J.D.
        • Herrenkohl T.I.
        Prevention in the school years.
        in: Farrington D.P. Coid J.W. Early Prevention of Adult Antisocial Behaviour. Cambridge University Press, Cambridge, UK2003: 265-291
        • Herrenkohl T.I.
        • Chung I.-J.
        • Catalano R.F.
        Review of research on predictors of youth violence and school-based and community-based prevention approaches.
        in: Allen-Meares P. Fraser M.W. Intervention with Children and Adolescents An Interdisciplinary Perspective. Allyn & Bacon, Boston, MA2004: 449-476
        • Fraser M.W.
        • Richman J.M.
        Resilience.
        in: Fraser M.W. Richman J.M. The Context of Youth Violence Resilience, Risk and Protection. Praeger, Westport, CT2001: 187-198
        • Wasserman G.A.
        • Miller L.S.
        The prevention of serious and violent juvenile offending.
        in: Loeber R. Farrington D.P. Serious and Violent Juvenile Offenders Risk Factors and Successful Interventions. Sage Publications, Inc, Thousand Oaks, CA1998: 197-247