Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
February 2004

Gun Carrying and Conduct Disorder: A Highly Combustible Combination?Implications for Juvenile Justice and Mental and Public Health

Author Affiliations

From the Departments of Psychiatry, University of Pittsburgh, Pittsburgh, Pa (Drs Loeber and Burke), and The University of Chicago, Chicago, Ill (Dr Lahey). Mr Mutchka is in private practice in Pittsburgh.


Copyright 2004 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2004

Arch Pediatr Adolesc Med. 2004;158(2):138-145. doi:10.1001/archpedi.158.2.138

Objectives  To examine concealed gun carrying between the ages of 12 and 17 years in a population of clinic-referred boys, many of whom qualified for a disruptive behavior disorder, including conduct disorder (CD); to identify factors and diagnoses related to concealed gun carrying; and to examine the extent to which gun carrying is associated with crime in adulthood.

Design  Longitudinal follow-up study.

Setting  Pittsburgh, Pa, and Athens and Atlanta, Ga.

Participants  One hundred seventy-seven clinic-referred boys, first assessed between the ages of 7 and 12 years and followed up yearly until the age of 19 years.

Main Outcome Measures  Violence, property offenses, and drug charges in adulthood.

Results  Between the ages of 12 and 17 years, 1 in 5 participants carried a concealed gun, and the annual prevalence increased linearly with age. More than half (61.1%) carried a gun for 1 year only. Gun carrying was significantly (incident rate ratio, 3.93%; 95% confidence interval, 1.60-9.60) associated with CD. Conduct disorder, maternal psychopathy, victimization, and parental monitoring increased the risk of gun carrying by a factor of 8. Adult crime was best predicted by gun carrying, CD, and parental monitoring. Gun carrying predicted drug charges, but not violence or property offenses.

Conclusions  Even though the carrying of handguns by juveniles is prohibited, young men with symptoms of CD are more likely to carry guns than young men without CD. The findings are discussed in terms of the need for the inclusion of gun carrying among the symptoms of CD.

Conduct disorder (CD) is a psychiatric diagnosis that is defined as a repetitive and persistent pattern of behavior in which the basic rights of others and major age-appropriate societal norms or rules are violated. The symptoms of CD include "has used a weapon that can cause serious physical harm,"1(p90) but do not mention the carrying of a concealed gun. Of all weapons used by juveniles, guns probably have the highest lethal impact. Yet, little is known about how many boys with CD illegally carry concealed guns before the age of 18 and to what extent their gun carrying is associated with violence. These issues are important because many CD boys have problems in controlling their impulsivity. Once these boys obtain a gun, they are potentially among the highest-risk groups to inflict violence on others or threaten others with violence. We see concealed gun carrying by CD boys as a problem for juvenile justice, a threat to public health, and a challenge for mental health professionals. (We will use the term gun carrying to mean concealed gun carrying.)

Several studies25 have linked delinquency, violence, and homicide rates to increased levels of the carrying of handguns in populations of males. Blumstein and Cork2 have shown that nationally, gun-related homicides by juveniles, in contrast to non–gun-related homicides by juveniles, have increased, starting in 1987. Longitudinal research4 shows that serious crime is higher during periods in which juveniles carry a gun.

Little is known, however, about the degree to which mental health problems, particularly CD, in childhood and adolescence increase the risk of gun carrying and violence. Bailey et al6 showed that students who brought a weapon to school were more likely to participate in fights and damage school property compared with students who did not bring a weapon to school. However, studies linking diagnoses in young populations to gun carrying seem to be lacking. Males with CD, probably because of their disruptive and aggressive behavior, are at high risk for illegally carrying a gun during adolescence. Several studies have emphasized that CD is overrepresented among the most serious delinquents, such as homicide offenders.7 Three quarters of homicides are committed with a handgun, which is similar to national figures.5 More is known about the predictors of violence, which also include alcohol use, early problem behavior, poor family relationships, and poor school bonding.810 It is not known, however, whether these and possibly other predictors apply to young men with CD who carry guns. Also, we do not know whether gun carrying during adolescence also increases the risk of delinquency during adulthood. We do not want to imply that gun carrying by CD boys must lead to their increased offending. The key issue, however, is whether gun carrying is associated with later criminality over and above what could be predicted based on earlier conduct problems alone. If there is a relationship between CD, gun carrying, and later delinquency, including violence, then this would have implications for juvenile justice, mental health services, and public health. Figure 1 summarizes the conceptual model and hypotheses governing this article.

Figure 1.
Image not available

Schematic diagram of the hypothesized model for gun carrying.

This article addresses the following questions: What is the prevalence of illegal concealed handgun carrying in a population of clinic-referred boys before the age of 18, and which psychiatric diagnoses are particularly associated with gun carrying? Which factors in their lives are associated with gun carrying? How delinquent and violent in adulthood are those with a psychiatric diagnosis in childhood, and is this related to earlier handgun carrying? And, does gun carrying mediate the risk of adverse criminal outcomes for CD boys in terms of violence, property offending, and drug charges in adulthood?


The data were collected as part of the Developmental Trends Study (Loeber et al11 provide details). One hundred seventy-seven clinically referred boys, aged 7 to 12 years at the initial assessment in 1987, were interviewed annually until the age of 19. Because of a funding cut in year 5, interviews were not administered. The average retention rate between years 2 and 13 was 91.9%, ranging from 97.7% in year 2 to 80.2% in year 12. Half (54.2%) of the sample was recruited from one site in Pennsylvania, and the other half from 2 sites in Georgia. Participants had to live with at least one biological parent and could not have a history of mental retardation or psychosis. Participants were 29.9% African American and 70.1% white. Approximately 52.5% of the participants lived in urban environments, 56.5% were not living with their biological father, and 40.7% came from families within the lowest 2 categories of the Hollingshead Four Factor Index of Social Status (A. B. Hollingshead, PhD, unpublished data, 1975). At the initial assessment, 38.4% of the participants met the criteria for Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition, CD, 65.0% for attention-deficit/hyperactivity disorder (ADHD), 34.5% for either overanxious disorder (OAD) or separation anxiety disorder (SAD), and 12.4% for either dysthymia or major depressive disorder.

Parent and child interviews were conducted until the child reached the age of 18 years; at the ages of 18 and 19 years, the child was interviewed alone, using measures designed for adults. Years 1 through 4 also included teacher interviews. No data were available from year 5 because of funding restrictions. The Institutional Review Board of the University of Pittsburgh gave approval and provided an annual review of the protocol. In all cases, parents and participants consented to participate.


Table 1 provides a summary of the constructs and measures used in the study. Within the drug charges construct, possession of a controlled substance, possession of a controlled substance device, possession with intent to deliver a controlled substance, and possession of drug paraphernalia were included. Alcohol-related charges were excluded. Because of the relatively few drug sales charges combined with the cross-sectional nature of this variable, we used the combined use and sales charges for the drug charges construct.

Table 1. 
Image not available
Constructs and Measures Used

The mean age of participants when criminally charged was 19.86 (SD, 1.93) years. Seventeen participants were charged with crimes before their age 17 assessment. Because we wanted to be predictive from gun carrying and its correlates to later crimes, we conducted the analyses with and without these participants. Their exclusion changed the results of only one model.


The outcome variable of gun carrying, as a panel variable between the ages of 13 and 17, was examined using a generalized estimating equation negative binomial model, with a logit link and robust estimator of variance. An exchangeable correlation structure was chosen based on the relatively small number of cases but many observations in the data set (Rabe-Hesketh and Everitt19 provide details). Grouping for all analyses was based on participant identification, and age was used to designate time.

Variables were tested for their independent relationship with gun carrying. Those found significant were then tested simultaneously within domains. To assess interactions, log likelihood tests were performed to compare models with and without the interaction term, and χ2 distributions of the differences in the models were used to determine the adequacy of the interaction in the model. Following a procedure outlined by Aiken and West,20 interactions were tested by transforming the variables so they were centered around their means. Those variables significant within the domain-level models were tested in a final model of predictors and correlates of gun carrying.

Variables retained as correlates of gun carrying, along with gun carrying itself, were tested as predictors of overall criminal charges in young adulthood and as individual categories of criminal charges. These models were negative binomial regressions, clustering on the participant.


Table 2 shows the prevalence of gun carrying and child psychiatric diagnoses by age. As is evident, the rates for all diagnoses decreased over time. At the age of 17, oppositional defiant disorder (ODD), CD, and ADHD were most prevalent, with 31.7% meeting the criteria for ODD, 19.5% for CD, and 17.7% for ADHD.

Table 2. 
Image not available
Prevalence Rates of Gun Carrying and Psychiatric Diagnoses by Age

Across the ages of 12 to 17 years, 36 (20.3%) of 177 participants carried a concealed gun at least once. Gun carrying started as early as the age of 10 and, as shown in Figure 2, there is an almost linear increase in gun carrying to the age of 17. By the age of 17, the prevalence of gun carrying had increased to approximately 12.2%; however, gun-carrying episodes varied greatly over time. Figure 3 shows that of all gun carriers, 61.1% reported carrying guns for only 1 year, 30.5% for 2 years, and only 8.4% for 3 to 4 years.

Figure 2.
Image not available

Prevalence of concealed gun carrying. The number of subjects for each age was as follows: 10 years, n = 60; 11 years, n = 63; 12 years, n = 76; 13 years, n = 146; 14 years, n = 145; 15 years, n = 141; 16 years, n = 133; and 17 years, n = 164.

Figure 3.
Image not available

Time spent gun carrying.

In bivariate regression models, gun carrying was significantly associated with CD (incident rate ratio [IRR], 3.93; 95% confidence interval [CI], 1.60-9.60), and was significantly and inversely related to OAD (IRR, 0.03; 95% CI, 0.01-0.10). Gun carrying was not significantly associated with ODD (P = .62), dysthymia (P = .48), depression (P = .53), SAD (P = .47), or ADHD (P = .34).


Bivariate analyses of correlates of concealed gun carrying within the domains of demographic characteristics, parental functioning, parenting, child functioning, substance use, and victimization were conducted. Significant correlates of gun carrying were as follows: adolescent violent behavior (IRR, 1.13; 95% CI, 1.12-1.14); the demographic variables of African American ethnicity (IRR, 2.68; 95% CI, 1.14-6.32), socioeconomic status (IRR, 0.60; 95% CI, 0.44-0.82), and urban residence (IRR, 4.45; 95% CI, 1.44-13.75); and the participant's age (IRR, 1.76; 95% CI, 1.40-2.22), such that each year increase in age was associated with a 76% increase in the rate of gun carrying. No significant interactions were found among the demographic variables. Regression analyses of socioeconomic status, ethnicity, and urban residence predicting gun carrying, controlling for age, showed that only socioeconomic status remained significant (P<.001).

Bivariate analyses using variables in the parental psychiatric disorders domain showed that scores on the maternal psychopathy index (IRR, 1.04; 95% CI, 1.01-1.09) were significantly correlated to the participant's rate of concealed gun carrying. Also, the victimization index (IRR, 0.91; 95% CI, 0.82-0.99) showed that as victimization scores increased there was a decrease in the rate of gun carrying. Within the parenting domain, only parental monitoring (inversely related) (IRR, 0.90; 95% CI, 0.84-0.96) was associated with gun carrying.

The final regression model of correlates of gun carrying, controlling for age (IRR, 1.69; 95% CI, 1.28-2.25), was as follows. The strongest correlates of gun carrying were adolescent violent behaviors (IRR, 1.07; 95% CI, 1.06-1.08) and CD (IRR, 5.24; 95% CI, 1.77-15.50), followed by victimization (inversely related) (IRR, 0.83; 95% CI, 0.75-0.93), maternal psychopathy (IRR, 1.06; 95% CI, 1.02-1.10), and parental monitoring (IRR, 0.86; 95% CI, 0.78-0.95). African American ethnicity, socioeconomic status, urban residence, and OAD were significant at the bivariate level, but were removed from the regression model. The following variables were not significant at the bivariate level: household income, marital status, maternal psychopathy, countercontrol, harsh discipline, marijuana use, alcohol use, hard drug use, tobacco use, and IQ. The main effects model indicates that there is greater than a 5-fold increase in the incident rate of gun carrying among males with CD compared with those without CD. In observations in which CD criteria were vs were not met, gun carrying occurred a mean of 7.43 (SD, 24.6) vs 1.66 (SD, 12.0) times. As parental monitoring increased by 1 U, the incident rate of gun carrying decreased by 14%. A significant (P = .02) interaction was found between victimization and maternal psychopathy in their effect on gun carrying (Figure 4). The interaction showed that for males with a high score on victimization, maternal psychopathy did little to influence gun carrying. On the other hand, for those with lower levels of reported victimization, higher scores of maternal psychopathy were associated with greater rates of gun carrying. In terms of the model, however, a likelihood ratio test showed that the interaction term did not significantly change the goodness of fit of the model; therefore, it was not retained in the model.

Figure 4.
Image not available

Interaction of victimization and maternal psychopathy on gun carrying. IRR indicates incident rate ratio.


The number of arrests per participant ranged from 0 to 16 (mean, 1.57; SD, 2.58). Nearly half (84 boys [47.5%]) of our sample was arrested at least once in adulthood, 55 (31.1%) had at least 2 arrests, and 37 (20.9%) had 3 or more arrests. First adult arrests occurred early, with 70.2% (59/84) of those with an adult criminal record having their first adult arrest by the age of 19. For those with an arrest, the number of charges ranged from 1 to 26 (mean, 7.04; SD, 6.12), and one quarter of those with an arrest had more than 11 charges. Regarding specific categories of charges, of those with any arrests, 41 (48.8%) were charged with violence, 49 (58.3%) with property crimes, and 43 (51.2%) with drug charges. Only 9 (10.7%) of 84 boys were charged with gun offenses, among whom was one participant charged with murder.

Conduct disorder (IRR, 2.26; 95% CI, 1.70-3.02) and ODD (IRR, 1.77; 95% CI, 1.21-2.58) were significantly related to the total number of charges. Overanxious disorder was also significant, but was inversely related to charges (IRR, 0.64; 95% CI, 0.43-0.97). Major depression was marginally, and inversely, related (IRR, 0.66; 95% CI, 0.43-1.04). The disorders of SAD (P = .58), dysthymia (P = .72), and ADHD (P = .14) were not significantly related to adult criminal charges. Regarding specific types of crimes, CD (IRR, 2.22; 95% CI, 1.27-3.88), ODD (IRR, 1.76; 95% CI, 1.02-3.06), and SAD (IRR, 1.60; 95% CI, 1.01-2.50) were predictive. Predicting drug charges, CD (IRR, 2.44; 95% CI, 1.48-3.99), ODD (IRR, 1.73; 95% CI, 1.01-2.97), and major depressive episodes (inversely related) (IRR, 0.34; 95% CI, 0.12-0.99) were significant. Overanxious disorder (IRR, 0.50; 95% CI, 0.21-1.17) was inversely and marginally associated with drug charges. Conduct disorder (IRR, 2.79; 95% CI, 1.94-4.03) and, marginally, ODD (IRR, 1.81; 95% CI, 0.93-3.51) and major depressive episodes (IRR, 0.59; 95% CI, 0.31-1.15) were associated with violent crimes.


To what degree will earlier gun carrying predict adult criminal outcomes when correlates of gun carrying are included? Predicting from the main effects model to criminal charges, gun carrying (IRR, 1.01; 95% CI, 1.00-1.01), CD (IRR, 1.97; 95% CI, 1.44-2.69), and parental monitoring (inversely related) (IRR, 0.95; 95% CI, 0.91-0.98) remained significant, while adolescent violent behavior, maternal psychopathy, victimization, and age (IRR, 0.98; 95% CI, 0.92-1.04) did not. (The data include cases in which one or more youth assessments were conducted after the first arrest. Exclusion of cases with such circumstances did not result in the addition or removal of any variables to this model.) A significant (P = .02) interaction was found between gun carrying and parental monitoring; high gun carrying in the presence of parental monitoring was associated with elevated rates of criminal outcomes. However, comparing the likelihood ratios of the models suggested that the interaction term did not contribute significantly to the overall model (χ2 = 1.4, P = .02) and was, therefore, not retained.

We then examined the prediction of specific categories of charges (Table 3). When predicting property crime charges, only CD remained significant. When predicting violent charges, victimization (inversely related), parental monitoring (inversely related), and CD remained significant, while gun carrying, maternal psychopathy, and age did not. Finally, the prediction of counts of drug charges showed significant main effects for gun carrying (IRR, 1.02; 95% CI, 1.01-1.03), parental monitoring (inversely related) (IRR, 0.90; 95% CI, 0.85-0.97), victimization (inversely related) (IRR, 0.92; 95% CI, 0.86-0.99), and CD (IRR, 2.11; 95% CI, 1.26-3.56). An interaction term involving gun carrying and parental monitoring (Figure 5) was significant (IRR, 1.00; 95% CI, 1.00-1.01), and improved the overall model (χ2 = 3.68, P = .05) (Table 3).

Table 3. 
Image not available
Predicting From the Gun-Carrying Model to Categories of Adult Criminal Charges*
Figure 5.
Image not available

Interaction of gun carrying by parental monitoring in the prediction of drug crimes. IRR indicates incident rate ratio.

Because some participants had criminal charges before the age of 17, we reran the analyses to examine whether this overlap had influenced the findings. Only in the case of the model predicting drug charges did their exclusion make a difference: victimization was no longer associated with drug charges when these participants were excluded from the analyses.


Finally, we wanted to examine whether gun carrying mediated the link between CD and adult criminal charges (Figure 1). Including gun carrying in models with CD as predictors of later adult crimes resulted in better-fitting models. In no cases did gun carrying fully mediate the relationship, and in the case of predicting property crimes and violent charges, gun carrying had no influence on the relationship, because it was not significant (P = .90 and .78, respectively) in the model. In the case of overall charges and of drug charges, specific statistical tests of mediation confirm that gun carrying serves as a partial mediator of the relationship (overall charges: Sobel test statistic = 2.46, P = .01; and drug charges: Sobel test statistic = 2.31, P = .02).


The study found that about 1 in 5 boys (20.3%) carried a concealed handgun before the age of 18. This compares with 18% in a national survey21 of 16 000 students in grades 9 through 12. However, the percentages are not entirely comparable because large school-based studies encompass a wider variety of youth compared with clinic-referred samples. Also, the results of cross-sectional population surveys are difficult to compare with populations followed up longitudinally from childhood to adulthood. The best we can say is that in the present study, gun carrying by this population of clinic-referred boys is modestly high.

Many of those carrying a gun did so at a young age. The youngest in this sample was at the age of 10; after that age, gun carrying increased linearly to a prevalence of 12.2% at the age of 17. Gun carrying was transitory: only a few gun-carrying boys (8.3%) carried a gun for 3 or more years, which is slightly higher than observed in the Rochester Youth Development Study.4

Of all psychiatric diagnoses, only CD was positively associated with gun carrying. This specific link probably rests on the association between delinquent-type symptoms of CD, reflecting a delinquent lifestyle, and gun carrying. Concealed gun carrying, as measured in this study, excluded the use of sporting guns. Conduct disorder, even when controlling for self-reported violent behavior, maternal psychopathy, victimization, and parental monitoring, increased the risk of gun carrying by a factor of 5.

In an earlier article22 based on data from the first wave of the present longitudinal study, the boys who met Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition, criteria for CD in wave 1 were divided into 2 groups based on whether they also met Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition, criteria for SAD or OAD. In wave 1, boys with CD and anxiety disorder were viewed as less aggressive, had fewer police contacts, and had been suspended from school less often than other boys with CD. We interpreted this finding in terms of the model of behavioral inhibition by Gray,23 regarding anxiety as an inhibitor of behavior in the presence of signals of punishment or nonreward. The present finding that OAD (but not SAD) is inversely related to gun carrying and the number of arrests through late adolescence is an important confirmation of the earlier finding. Future studies of processes underlying OAD, or generalized anxiety disorder, that potentially inhibit serious conduct problems could be highly important to psychological models of antisocial behavior.

Almost half of the boys in the study (47.5%) had been arrested as adults, and half of those with any charges were charged with a violent crime. However, only 1 in 10 of those charged were charged with gun offenses. Adult crime was best predicted by gun carrying, CD, and parental monitoring. Also, for males who had lower levels of reported victimization, higher maternal psychopathy was associated with greater rates of gun carrying. Gun carrying predicted drug charges, but not violence or property offenses. Gun carrying, over and above CD, predicted later crime, particularly drug charges. Because of relatively low counts of drug dealing, we could not clearly distinguish between drug charges for possession or drug dealing. However, we believe that there is a greater link between gun carrying and drug dealing than between gun carrying and possession of drugs, because of the inherent dangers of victimization and robbery associated with drug dealing rather than use or possession. Nevertheless, we cannot exclude the possibility that drug possession identified some who were dealers as well. This point should be addressed in future studies.

In summary, the results show that several factors associated with gun carrying in general populations also applied to this clinic-referred sample. What is new is that CD, which often can be observed earlier than patterns of delinquent acts, was uniquely predictive of gun carrying.

The study has several limitations. It consists of a relatively small clinically referred sample, which may limit the generalizability of the findings, and measures of self-reported delinquency and gun carrying began in year 3. Even though self-reports of gun carrying are used in survey studies,24 the degree to which it is underreported or overreported remains unclear. On the positive side, self-reports of gun carrying have predictive utility, as shown in this study.

Even though the carrying of handguns by juveniles is prohibited, the pediatric and psychiatric literature has not sufficiently addressed the extent to which youth with symptoms of CD are likely to carry guns. We make the case that gun carrying, even if it is transitory and independent of CD, is predictive of adult crime, particularly drug charges, in a clinical sample of young men. If replicated, these findings would support the inclusion of concealed gun carrying as a symptom of CD in a future revision of Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition. We also argue that preventing youth who qualify for CD from carrying concealed guns should be of the highest priority. We realize, however, that such prohibitions are only as effective as enforcement can be put in place. We recommend that a dialogue be started between professional organizations representing pediatricians, psychiatrists, and other mental health workers and local police and juvenile justice personnel to target gun-carrying CD youth as a method to reduce and prevent crime in communities.

What This Study Adds

Even though concealed gun carrying is not a symptom of CD, it has the potential of inflicting lethal harm. Studies on gun carrying by children and adolescents who qualify for CD seem unavailable. It is unclear to what extent gun carrying predicts adult crime and whether CD carries additional risk. These issues were examined in a sample of clinic-referred boys, who were followed up into adulthood. Results show that between the ages of 12 and 17, 1 in 5 of the young men carried a concealed gun, which linearly increased with age. More than half (61.1%) of the young men carried a gun for 1 year only. Gun carrying was significantly associated with CD. Conduct disorder, maternal psychopathy, victimization, and parental monitoring increased the risk of gun carrying by a factor of 8. Adult crime was best predicted by gun carrying, CD, and parental monitoring. Gun carrying predicted drug charges, but not violence or property offenses. Implications are discussed for clinical assessments and for the future symptom range of the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition.

Back to top
Article Information

Corresponding author: Rolf Loeber, PhD, Western Psychiatric Institute and Clinic, Department of Psychiatry, University of Pittsburgh, 3811 O'Hara St, Pittsburgh, PA 15213-2593 (e-mail:

Accepted for publication August 22, 2003.

This study was supported by grant MH42529 from the National Institute of Mental Health, Rockville, Md.

American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition.  Washington, DC American Psychiatric Association1994;
Blumstein  ACork  D Linking gun availability to youth gun violence. Law Contemp Probl. 1996;591- 24Article
Cooke  PJLaub  JH Recent trends in youth violence in the United States. Tonry  Med.Youth Violence. Chicago, Ill University of Chicago Press1998;27- 64
Lizotte  ASheppard  D Gun Use by Male Juveniles: Research and Prevention.  Washington, DC US Dept of Justice2001;OJJDP Juvenile Justice Bulletin
Snyder  HNSickmund  M Juvenile Offenders and Victims: 1999 National Report.  Washington, DC Office of Juvenile Justice and Delinquency Prevention, US Dept of Justice1999;
Bailey  SLFlewelling  RLRosenbaum  DP Characteristics of students who bring weapons to school. J Adolesc Health. 1997;20261- 270
Heckel  RSchumaker  DM Children Who Murder.  Westport, Conn Praeger Publishers2001;
Ellickson  PLMcGuigan  KA Early predictors of adolescent violence. Am J Public Health. 2000;90566- 572
Hawkins  JDHerrenkohl  TFarrington  DPBrewer  DCatalano  RFHarachi  TW A review of predictors of youth violence. Loeber  RFarrington  DPeds.Serious and Violent Juvenile Offenders: Risk Factors and Successful Interventions. Thousand Oaks, Calif Sage Publications1998;106- 146
Lipsey  MWDerzon  JH Predictors of violent or serious delinquency in adolescence and early adulthood: a synthesis of longitudinal research. Loeber  RFarrington  DPeds.Serious and Violent Juvenile Offenders: Risk Factors and Successful Interventions. Thousand Oaks, Calif Sage Publications1998;313- 345
Loeber  RGreen  SMLahey  BBFrick  PJMcBurnett  K Findings on disruptive behavior disorders from the first decade of the Developmental Trends Study. Clin Child Fam Psychol Rev. 2000;337- 60
Elliott  DSHuizinga  DAgeton  SS Explaining Delinquency and Drug Use.  Beverly Hills, Calif Sage1985;
Thornberry  TPHuizinga  DLoeber  R The prevention of serious delinquency and violence: implications from the Program of Research on the Causes and Correlates of Delinquency. Howell  JCKrisberg  BHawkins  JDWilson  JJeds.Sourcebook on Serious, Violent and Chronic Juvenile Offenders. Thousand Oaks, Calif Sage Publications1995;213- 237
Costello  AJEdelbrock  CDulcan  MKKalas  RKlaric  S Diagnostic Interview Schedule for Children (DISC).  Pittsburgh, Pa Western Psychiatric Institute and Clinic, School of Medicine, University of Pittsburgh1987;
Wechsler  D Manual for the Wechsler Intelligence Scale for Children–Revised.  San Antonio, Tex Psychological Corp1974;
Hathaway  SRMcKinley  JC The Minnesota Multiphasic Personality Inventory.  Minneapolis University of Minnesota Press1943;
Spitzer  RLWilliams  JGibbon  M Structured Clinical Interview for DSM-III-R: Nonpatient Version–NP.  New York New York State Psychiatric Hospital1987;
Loeber  RFarrington  DPStouthamer-Loeber  MVan Kammen  WB Antisocial Behavior and Mental Health Problems: Explanatory Factors in Childhood and Adolescence.  Mahwah, NJ Lawrence Erlbaum1998;
Rabe-Hesketh  SEveritt  BS A Handbook of Statistical Analyses Using Stata. 2nd ed. Boca Raton, Fla Chapman & Hall/CRC2000;
Aiken  LSWest  SG Multiple Regression: Testing and Interpreting Interactions.  Newbury Park, Calif Sage1991;
Kann  LKinchen  SAWilliams  BI  et al.  Youth risk behavior surveillance: United States, 1997. MMWR CDC Surveill Summ. 1998;471- 89
Walker  JLLahey  BBRusso  MF  et al.  Anxiety, inhibition, and conduct disorder in children, I: relations to social impairment. J Am Acad Child Adolesc Psychiatry. 1991;30187- 191
Gray  JA The Neuropsychology of Anxiety: An Enquiry Into the Functions of the Septo-Hippocampal System.  Oxford, England Oxford University Press1982;
Puzzanchera  CM Self-reported delinquency by 12-year olds, 1997.  Washington, DC US Dept of Justice, Office of Justice Programs, Office of Juvenile Justice and Delinquency Prevention2000;1- 2OJJDP Fact Sheet, No. 3