1 GENDER SPECIFIC AND COMMON DEVELOPMENTAL TRAJECTORIES OF AGGRESSION, DELINQUENCY, AND SUBST ANCE USE ACROSS MIDDLE SCHOOL: THE ROLE OF DEVIANT PEER ASSOCIATION AND SENSATION SEEKING By SARAH DELPHIA LYNNE A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2008
2 2008 Sarah Delphia Lynne
3 To my mother, Karen, my brothe r and sister, Jonathan and Elyse, and my ever supportive husband, Matthew
4 ACKNOWLEDGMENTS There are several people who have played s upportive roles for me during my years as a graduate student. I would like to take this oppo rtunity to thank these pe ople for their guidance and encouragement during this formative experience in my life. First, I would like to express my most sincere gratitude to my advisor, Dr. Julia Graber. Th roughout my graduate training she has provided exceptional insight, encouragement, and support, aiding in the development and refinement of my research inte rests in promoting health and re ducing negative adjustment during the adolescent transition. She is a tremendous academic who has served as a prime example to me of a strong, intelligent, and successful woma n. Her expertise and guidance have provided me with the skills and confidence to pursue my goals of continued academic research and excellence in the area of adolescent adjustment. Special thanks go to my committee member s, Drs, Lise Youngblade, Alex Piquero, James Algina, and James Shepperd. With thei r help, I have learned the importance of thoroughness and clarity in conducti ng and reporting scientific resear ch and I truly feel confident in my ability to do so. I would also like to offer thanks to my fellow graduate students in the Developmental Psychology department for c ontinually offering unconditional support and friendship. Without the support of my friends, my graduate school experience would have been significantly le ss enjoyable. I am also very thankful to my family. The support and encouragement that my mother, brother and sister, and my grandpa rents have given me has been invaluable. I thank my mother for showing me, by example, that I can achieve any goal I set for myself through hard work, patience, and perseverance. It is my family s unconditional and unwavering support, love, and encouragement that has sustained me and helped me to succeed.
5 Finally, I am extremely thankful for the support and love of my husband, Matthew. He has been a limitless source of strength and sola ce for me. His patience and understanding during the most stressful times of my graduate training we re invaluable in helping me to persevere. I know that he will continue to support me and all of my goals in th e years to come.
6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ............................................................................................................... 4LIST OF TABLES ...........................................................................................................................9LIST OF FIGURES .......................................................................................................................10ABSTRACT ...................................................................................................................... .............11CHAPTER 1 A DEVELOPMENTAL MODEL OF RECIPROCAL INFLUENCES AMONG ANTISOCIAL BEHAVIORS ................................................................................................13Definition and Assessment of Antisocial Behaviors ..............................................................16Aggression .......................................................................................................................17Delinquency .....................................................................................................................19Substance Use ................................................................................................................. .20Rates of Aggression, Delinquency, and Substance Use in Adolescence ................................22Aggression and Delinquency ........................................................................................... 23Substance Use ................................................................................................................. .25Developmental Models of Aggressi on, Delinquency, and Substance Use .............................27General Developmental Models ...................................................................................... 28Outcome Specific Models ............................................................................................... 29Models Applicable to Several Antisocial Behaviors ....................................................... 31Models of Associations between Outcomes .................................................................... 35The Developmental Model of Reciprocal Influence .............................................................. 37Early Adolescence ...........................................................................................................44Late Adolescence and the Tr ansition to Adulthood ........................................................ 48Reciprocal Influences ......................................................................................................49Salient Issues for Understandi ng Reciprocal Influences ................................................. 50Conclusions .............................................................................................................................54The Proposed Study ................................................................................................................572 THE ROLE OF SENSATION SEEKING AND DEVIANT PEER ASS OCIATION ON ANTISOCIAL BEHAVIOR ................................................................................................... 59Pathways to Drug Use and Aggression a nd Delinquency: Conceptual Framework .............59Core Domains .................................................................................................................. .......62Individual Sensation Seeking ..........................................................................................62Peer Factors .....................................................................................................................653 SPECIFIC AIMS ................................................................................................................. ...704 RESEARCH DESIGN AND METHODS .............................................................................. 73
7 Design ........................................................................................................................ .............73Participants .................................................................................................................. ....73Procedure ..................................................................................................................... ....74Measures ...................................................................................................................... ...........75Demographics .................................................................................................................. 75Drug Use ..........................................................................................................................76Aggression .......................................................................................................................76Delinquency .....................................................................................................................77Friend Delinquency .........................................................................................................78Friend Drug Use ..............................................................................................................78Sensation Seeking ............................................................................................................79Tracking and Attrition ............................................................................................................80Analysis Plan ..........................................................................................................................81Power Analysis ................................................................................................................81Longitudinal Analyses .....................................................................................................82Testing Mediating and Moderating Effects .....................................................................82Analysis of Specific Aims ...............................................................................................835 RESULTS ....................................................................................................................... ........87Descriptive Analysis .......................................................................................................... .....87Group-Based Trajectory Analyses ..........................................................................................93Aggression .......................................................................................................................94Delinquency .....................................................................................................................97Substance Use ................................................................................................................ 100Dual Trajectory Analysis ......................................................................................................103Aggression and Delinquency ......................................................................................... 104Aggression and Substance Use ...................................................................................... 105Delinquency and Substance Use ...................................................................................107Mediation Analyses ..............................................................................................................109Aggression and Delinquency as Me diators of Sensation Seeking ................................110Aggression and Delinquency as Me diators of Friend Delinquency ..............................111The Moderating Role of Gender ........................................................................................... 112Sensation Seeking by Gender ........................................................................................113Friend Delinquency by Gender ..................................................................................... 114The Moderating Role of Race / Ethnicity ............................................................................. 1166 DISCUSSION .................................................................................................................... ...121Strengths and Implications of Study Design ........................................................................ 129Limitations ................................................................................................................... ..130Implications .................................................................................................................. .132APPENDIX LIFE SKILLS TRAINING HEALTH SURVEY ........................................................................ 135
8 LIST OF REFERENCES .............................................................................................................138BIOGRAPHICAL SKETCH .......................................................................................................148
9 LIST OF TABLES Table page 1-1 Summary of studies that exam ined reciprocal e ffects among antisocial behaviors ...........415-1 Descriptive statistics .................................................................................................... ......875-2 Correlations among the predictor variables .......................................................................915-3 Correlations among outcome variables .............................................................................. 925-4 Using BIC to select the number of gr oups to include in the aggression model ................. 955-5 Descriptive statistics with in aggression trajectory group ..................................................965-7 Using BIC to select the number of groups to include in the substance use model .......... 1015-8 Descriptive Statistics within Substance Use Trajectory Group ....................................... 1035-9 Probability of delinquency trajectory group membership given aggression trajectory group membership ...........................................................................................................1045-10 Probability of substance use trajectory group membership given aggression trajectory group membership ...........................................................................................................1065-11 Probability of substance use trajectory group membership given delinquency trajectory group membership ...........................................................................................1075-12 Hierarchical linear models examining the association between gender, sensation seeking, and substance use ...............................................................................................1135-13 Hierarchical linear models examini ng the association between gender, friend delinquency, and delinquency ..........................................................................................1155-14 Hierarchical linear models examining th e association between Latino race/ethnicity, sensation seeking, friend de linquency, and aggression ................................................... 117
10 LIST OF FIGURES Figure page 1-1 The developmental model of reciprocal influence .............................................................383-1 Hypothetical model of individual an d contextual pathways to drug use ........................... 615-1 Five group aggression trajectory model ............................................................................. 965-2 Six group delinquency trajectory model ............................................................................ 995-6 Mediation model evaluating aggression as the mediator ................................................. 1095-4 Gender moderating the association betw een person-centered sensation seeking and substance use ....................................................................................................................1145-5 Gender moderating the association betw een person-centered friend delinquency and delinquency ................................................................................................................... ...1165-7 A three-way interaction between Latino race/ethnicity, sens ation seeking, and delinquent peer association on aggressive behavior ........................................................119
11 Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GENDER SPECIFIC AND COMMON DEVELOPMENTAL TRAJECTORIES OF AGGRESSION, DELINQUENCY, AND SUBST ANCE USE ACROSS MIDDLE SCHOOL: THE ROLE OF DEVIANT PEER ASSOCIATION AND SENSATION SEEKING By Sarah Delphia Lynne May 2008 Chair: Julia A. Graber Major: Psychology During adolescence, increases are observed in rates of aggression, delinquency, and drug use with strong evidence of diffe rences between males and females. Future prevention efforts would be greatly aided by a better understanding of gender specific and common pathways to drug use that incorporate interconnections with pathways to a ggression/delinquency. The current study evaluated bi-directiona l and temporal associati ons between drug use and aggression/delinquency across 6th, 7th, and 8th grades using data available from the control group of an evaluation of a drug use and violence prev ention program. Group-based trajectory analysis revealed trajectories of aggression, delinquency, and substance use which support the existence of both adolescent-limited and life-course persis tent offenders. In addition, unexpected patterns of increasing/decreasing change were observe d during middle school for both aggression and delinquency. Clear temporal associations we re observed between deve lopmental changes in aggression, delinquency, and substance use. There were few gender differences in the developmental progression of these problem be haviors during middle school with only two exceptions, males were more like to follow trajec tories of increasing/decreasing aggression and high stable delinquency. Evalua tions of ethnic/racial differe nces in the trajectory group
12 membership revealed that significantly more Afri can American adolescents followed trajectories of increasing/decreasing, high increasing, and chronic high aggr ession and delinquency compared to Latino and White/other adolescents. Hierarchical linear modeling revealed a more pronounced influence of sensation seeking and de viant peer association on the development of antisocial behaviors among females compared to males. In addition, individual changes in sensation seeking conferred a st ronger risk for aggression am ong Latinos compared to other race/ethnicities. Changes in association with delinquent peers were less influential for Latinos compared to other race/ethnicities. Finally, associations between sensation seeking and delinquent peer association on substance use were partially mediated by individual changes in both aggressive and delinquent behavior. The resu lts of this study provide important information regarding interconnect ions between developmental changes in antisocial behavior that occur during the middle school years. The knowledge gained from this study regarding individual and contextual factors and their connection to path ways for drug use and aggression/delinquency will inform etiology of drug use, drug prevention co ntent, and program implementation strategies.
13 CHAPTER 1 A DEVELOPMENTAL MODEL OF RECIPROC AL INFLUENCES AMONG ANTISOCIAL BEHAVIORS The etiology and prevention of adolescent s ubstance initiation and subsequent use and abuse have been the focus of extensive research and evaluation. The liter ature on this topic is broad and varies by how substance use is conceptualized. Reviews on the topic of substance use have covered such aspects as the distinction between use and abuse (Newcomb & Bentler, 1989), multiple risk and protective factors over the lifespan (Mayes & Suchman, 2006; Newcomb, 1997), and gender differences in substance abus e (Andrews, 2005). In addition, some studies focus on correlates and developmental pathways fo r specific substances (e.g., alcohol or tobacco) whereas others examine any or multiple substance initiation and use. Moreover, it has been widely acknowledged that the pr evention and treatment of adoles cent drug use benefits from consideration of other types of antisocial behaviors in adoles cence (Newcomb, 1997). Notably, the development of aggression and delinquency across childhood and adolescence has also been the focus of extensive evaluation. Numerous books, chapters, and review articles have been published on aggression and delinquency reflecting definitional issues, the emergence of these behaviors, correlates of aggression and de linquency, and prevention efforts (Dishion & Patterson, 2006; Dodge, Coie, & Lynam, 2006; Moff itt, 2006; Moffitt, Caspi, Rutter, & Silva, 2001). The purpose of this review is to integrat e research findings regarding interconnections between substance use and other antisocial beha viors such as aggressi on and delinquency with particular attention to the role of gender and ethnicity. Given th e large number of studies that have been reported on each of these topics indivi dually, this review focuses on only those studies which have evaluated the pred ictive influence of aggression or delinquency on substance use during adolescence and early adulthood, or the predictive influence of substance use on aggression and delinquency duri ng this same time period.
14 Several widely accepted conclusions have been drawn about antisocia l behaviors such as aggression, delinquency, and substance use (P iquero, Farrington, & Blumstein, 2007). The current paper focuses on one of these conclusions: Different types of antisocial behaviors begin to emerge at distinctively different ages. For in stance, aggressive behavior in childhood tends to precede the onset of delinquent be havior which more often increas es in adolescence, delinquent behavior in early adolescence tends to precede s ubstance use in mid to late adolescence, and so forth. Despite this relatively well established conclusion, there are still areas which require further inquiry to clarify points of contention within the field. While certa in antisocial behaviors have clearly been shown to preced e other behaviors in a predictable sequence, it is unclear if this is due to developmentally appropriate changes in the manifestation of an underlying construct of general deviance, or if the onset of one behavior serves as an impetus or stepping stone towards engagement in future antisocial behaviors. In particular, very lit tle is known about how antisocial behavior in childhood may serve as a stepping stone to engagement in delinquent behavior and substance use in ad olescence and how this relates to adulthood criminal offending. This leads to a number of que stions regarding the developm ental sequencing of antisocial behaviors. What is the relati onship between past engagement in antisocial behavior and future antisocial behaviors? Does the pr edictable sequencing in onset of these behaviors reflect specific causal associations among behavior s? Or is it reflective of pe rsistent individual differences which are manifested as different antisocial behaviors dependi ng on developmental stage? Or does the predictable sequencing of antisocial beha viors arise due to bi-directional associations between an individuals past behavior and their personal characteristics? A great deal of research has been published re garding risk and protec tive factors associated with aggression, delinquency, and substance use i ndividually. In addition, a variety of theories
15 have been developed addressing the co-occurren ce of antisocial behaviors as either stemming from an underlying construct of general devi ance (Gottfredson & Hi rschi, 1990) or from common risk factors associated with the developm ent of multiple antisocial behaviors (Jessor, 1992). These theories have inspired a large body of research, a subset of which has evaluated engagement in aggressive or delinquent activit ies as predictive of substance use in early adolescence, the time during which onset of delin quent behavior and substance use often occurs (Farrell, Sullivan, Esposito, Meyer, & Valois, 2005; Scheier & Botvin, 1996). A separate body of research has evaluated substance use as pred ictive of engagement in delinquent and criminal behavior in late adolescence and adult hood (Stacy & Newcomb, 1995). There is also international interest in reciprocal influen ces between antisocial behaviors; however, the literature is sparse (Laventure, Dry, & Pauz, 2006). The present disc ussion is not meant to serve as a comprehensive review of the descriptive and etiological studies of these behaviors. However, previous models of antisocial behavior have not attempted to explain the reciprocal influences between aggression, delinquency, a nd substance use implied by these lines of research. The present paper proposes a Developm ental Model of Recipro cal Influences (DMRI) which integrates and expands upon the interconne ctions between aggression, delinquency, and substance use across time; specif ically modeling both causal asso ciations among these antisocial behaviors as well as interindividual differences in risks factors. The goals in presenting the DMRI are to (a) integrate findings from disparate areas of antisocial behavior research (e.g., criminology, developmental psychology); (b) offer a model capable of explaining developmental sequen ces in onset of aggr ession, delinquency, and substance use from late childhood through adulthood th at is consistent with current theories of antisocial behavior; and (c) expl icate particular mechanisms contributing to the sequential
16 emergence of aggression, delinquency, and s ubstance use including both interconnections between these antisocial behaviors as well as in terindividual differences associated with the development of these antisocial behaviors. In so doing, we offer a framework through which future research can clarify the mechanisms underlying the developmental sequencing of aggression, delinquency, and substance use from childhood through adulthood. We begin by discussing construct definitions and assessment issues followed by a section on current prevalence rates of aggression, delinque ncy, and substance use. We then provide a brief discussion of current models and theories which address risk factors associated with the onset of these behaviors individually as well as associations among these antisocial behaviors. The following section introduces the Developmen tal Model of Reciprocal Influences (DMRI) among aggression, delinquency, and substance use. Th is section directly addresses the influence of an individuals past behavior on their future behavior and presen ts research that supports this model. We also discuss factors th at seem most salient to testi ng reciprocal influences both in terms of constructs (e.g., gender and ethnicity) and methodolo gy (e.g., analytic approaches). Finally, the concluding section makes recommen dations and highlights important issues to consider in the development of future studie s of aggression, delinquency, and substance use, including the major limitations of this type of research and suggesti ons for addressing these limitations. Implications for prevention scientis ts, policy makers, and developmental researchers are addressed. Definition and Assessment of Antisocial Behaviors Prior to discussing developmental models of aggression, delinquency, and substance use, it is important to note that different models will use different terms to refer to groups of antisocial behaviors (e.g., reckless behavior, de viant behavior, problem behavior, etc.). In fact, the lack of consensus regarding terminology in this area of inquiry creates a challeng e for the interpretation
17 and integration of results across studies. For th e purposes of clarity, th roughout this review the term antisocial behavior refers to engagement in aggressive, delinque nt, or substance using behaviors. When discussing models which use different terminology, definitions will be provided to clarify what behavior s are encompassed by such terms. In the present section, we discuss in more detail issues surrounding the definition and measurement of antisocial behaviors. Aggression The term aggression is commonly applied to act s intended to cause harm to others (Dodge et al., 2006; Parke & Slaby, 1983). Initial evaluations of aggre ssion often did not distinguish between different forms of aggressive behavior. Physically aggressive behavior received the most attention until Crick and Grotpeter (1995) introduced the concept of relational aggression. While physical aggression results from the intent to physically harm another individual, relational aggression is the intent to harm via social mechanisms such as exclusion from the group or spreading rumors. Distinct from, but often co-occurring with, physical and relational aggression, is verbal aggression. Verbal aggression and hostility includes threats or insults aimed at harming another individual. In addi tion, aggression has been distinguished by the motive of the aggressor with in strumental aggression seen as a means to an end and hostile aggression seen as an emotional reacti on (Hartup, 1974). A c onceptually similar subclassification of aggression is proactive versus reactive aggr ession. Proactive aggression is similar to instrumental aggression in that the individual aggresses in an ticipation of attaining some self-serving goal. Reactive aggression is similar to hostile aggression in that it occurs in response to an antecedent such as provocation (Dodge et al., 2006). In more extreme cases of chronic aggressive behavior, ch ildren are diagnosed with psyc hiatric conditions such as oppositional defiant disorder or conduct disorder. While these children are at the highest risk for
18 future violence and delinquency, sub-clinical levels of oppositional behavior during childhood are also a pathway to future aggression and delinquency (Loeber & Hay, 1997). In research on aggression, multiple forms of assessment have been used. In childhood, aggressive behavior has been quantified through natura listic observation of children interacting with one another. Participants have also been placed in experimental conditions meant to illicit aggressive responses (Dodge et al., 2006). Simila rly, there are adolescent activities and vignettes employed which are meant to pot entially illicit aggressive responses. Each of these methodological designs utilizes a coding format with which behaviors are categorized according to degree of aggressive behavior. Along with these more objectively observable measures of aggression, perceptions of aggressi ve behavior are often measured via parent, teacher, and self report as these are the indi viduals most likely to be aware of a childs or a dolescents behaviors. Evaluations of consistency between informants re veal low but significant associations (average r s ranging from .22 to .28) indicating that different informants present unique perspectives on child/adolescent/self behavioral and emoti onal problems (Achenback, McConaughy, & Howell, 1987; Phares, Compas, & Howell, 1989). Consiste ncy between informants is slightly higher among children (up to age 11) rather than adolescents (ages 12 19). Typically, adolescents, in comparison to children, have more opportunities to engage in aggressive behaviors that are not observed by parents or teachers. Due to the variety of ways that aggression has been conceptualized and measured across studies, it is important to be cautious when attempting to gene ralize effects. Consideration should be paid not only to the operational definition of aggression used but also to characteristics of the participants in the study. While large numbers of studies have evaluated aggression within the normal population, a substantial portion of re search on aggression has involved children and
19 adolescents from clinical populations, such as thos e diagnosed with conduct disorder. Studies of these individuals provide necessary information about the most severe forms of aggressive behavior. However, results from such studies may not generalize to individuals within the subclinical range of aggressive behavior. Delinquency Delinquent behavior is typically defined as any behavior which is considered a criminal offense if committed by an adult (Dishion & Pa tterson, 2006). In the United states, status offenses have also been included under the rubr ic of delinquency, as th ey refer to specific behaviors which are considered criminal offenses if committed by a minor (e.g., truancy or alcohol use). However, most Eu ropean countries consider status offenses to be a form of antisocial behavior rather than a criminal o ffense (Junger-Tas, Marsha ll, & Ribeaud, 2003). Hence, some assessments of delinquent behavior in adolescence include substance use items while others do not and caution shou ld be taken when evaluating studies of delinquent behavior. However, studies that include both substance use and delinqu ency as separate outcomes consistently remove all substance use items from the assessment of delinquency to reduce multicolinearity. Delinquency in adolescence is most frequent ly evaluated via adol escent self report or government records of criminal offenses. Typi cally, self report provides a more complete picture of rates of delinquent behavior. Through out adolescence, 15% 33% of individuals who commit delinquent acts are arrested. Of those arrested, only 64% are referred to court where between 2% and 10% are ultimately convicted (D odge et al., 2006). Hence, those individuals involved in the criminal justice system duri ng adolescence represent a specific subset of adolescents who engage in more delinquent behavior in general. This results in a comparable sample selection bias as when us ing clinical samples of adolescen ts to evaluate aggression. In
20 addition, first arrest often occurs years after an individual first begins engaging in delinquent behavior. Therefore, capturing the initial onset and escalation of delinquent behavior requires self-report. Research on delinquency has evaluated this construct in general as well as violent and non-violent subtypes. Despite the high correlation between violent and non-violent delinquency, the distinction between these forms of delinquent behavior merit separate analysis due to the real world implications associated with each. Substance Use As with aggression and delinque ncy, studies that evaluate substance use define and measure this construct in a vari ety of different ways making it difficult to summarize all of the known literature on the topic. As noted, distinctions are made between initiation of substance use, frequency of substance use, and distinguis hing substance use from substance abuse. Some studies evaluate single substances such as tobacco use or alco hol use only. Others combine multiple substances into a singl e measure of overall substance use (e.g., polydrug use) while still others combine substance use with other an tisocial behaviors in cluding aggression and delinquency into a construct of general deviance. While a few studies have experimentally investigated tobacco and alcohol use in samples of college students and adults there are obvious ethical re strictions on experimental manipulation of substance use with adolescent s or children (Fischman & Johanson, 1998). Hence, this construct is most fr equently evaluated via self-repor t. This, of course, introduces bias resulting in either underreporting or exagge rating of substance use, and antisocial behaviors more globally (McCord, 1990). A number of techni ques have been developed to attenuate this bias. First, confidentiality of participant respon ses is frequently discussed with the adolescent before they begin responding to the survey and random identification codes are often used in
21 place of names or other identifying information. In addition, the bogus pipeline procedure has also been developed to increase the accuracy of reports of tobacco use (Evans, Hansen, and Mittelmark, 1977). In this procedure, adolescent s provide a breath sample which they are told will be analyzed for levels of carbon monoxide to verify the accuracy of their reports of cigarette smoking behaviors. In fact, thes e samples are not analyzed but th is procedure has been shown to increase the validity of self reported cigarett e smoking as well as other antisocial behaviors (Tourangeau, Smith, & Rasinski, 1997). Similarly, st udies have also utilized saliva samples for information regarding cigarette use. Samples are analyzed to determine levels of carbon monoxide and thiocyanate. However, this technique has not been es tablished as a valid alternative to self-reported t obacco use, as these methods prove unreliable with adolescent populations (Snow, Gilchrist, & Schinke, 1985). A methodological technique used to increase the validity of self reported substance use involves the format of the questi ons asked. Often individuals ar e reluctant to discuss current substance use as often times it is frowned upon by so ciety and may even invol ve illegal behavior. To avoid this situation, substan ce use questions are often phrased to refer to past events (Day, Wagener, & Taylor, 1985). This allows participan ts to more honestly discuss their past behavior without indicating that they necessarily plan on continuing that behavior in the future. In addition, changes in survey format have also increased the validity of self reported antisocial behaviors. Computer based survey fo rmats, as opposed to paper and pencil surveys or interviews, have been shown to yield greater reports of engageme nt in antisocial behaviors such as substance use, delinquent behavior, and se xual behavior (Booth-Ke wley, Larson, & Miyoshi, 2007; Turner, Ku, Rogers, Lindberg, & Pleck, 1998). The greater reports of antisocial behaviors that result from computer based surveys are hypothes ized to be more accurate due to participants
22 increased comfort answering sens itive questions as a result of the impersonal context of the computer. Rates of Aggression, Delinquency, and Substance Use in Adolescence Given societal concerns about adolescent engagement in antisocial behaviors, two large national survey studies monitor rates of these behaviors on a regular basis. No comparable databases exist for the childhood years; hence, behavioral frequencie s in childhood are drawn from recent reviews. Nationally representative historical records of rates of adolescent health-related behaviors during high school have been maintained bienni ally since 1991 via the Youth Risk Behavior Surveillance System (YRBSS). The YRBSS includes extensive records of s ubstance use, as well as more serious forms of aggression and delinqu ency. The YRBSS is aimed at evaluating the historical prevalence of antisocial behavior s (e.g., aggression, deli nquency, substance use) among high school students and how these beha viors are distributed among subgroups of adolescents (CDC, 2006). Monitoring the Future (MTF) is a nationally representative survey of the beliefs, attitudes, and behaviors of adolescents with an emphasis on monitoring adolescent alcohol and other substance use (Johnston, OMalley, Bachman, & Schul enberg, 2006). This project has surveyed high school seniors since 1975 and has included eighth grade and te nth grade students annually since 1991 with a subsample of youth completing fo llow-up surveys at later points in time. However, given that the results of MTF focus h eavily on substance use, the rates reported below will focus on results from the YRBSS. Previous research has indicated that males te nd to engage in higher rates of aggression, delinquency, and substance use compared to females (Dodge et al., 2006; Mayes & Suchman, 2006). In addition, differences have been observed between ethnic and racial groups such that
23 White adolescents report higher ra tes of substance use and lower rates of delinquent behavior compared to Black adolescents (Johnston et al., 2006; Mayes & Suchman, 2006). Given the previous findings of gender and race/ethnicity di fferences, it is important to discuss changes in rates of antisocial behaviors se parately for these subgroups in addition to trends over time. Aggression and Delinquency Developmental increases in fr equency of aggressive beha vior peak in toddlerhood; however, aggressive behavior in adolescence and early adulth ood is more dangerous to both victims and perpetrators (Dodge et al., 2006). Increases in bot h aggression and delinquency occur in early adolescence (age s 11 14) peaking in late adol escence and early adulthood. Most commonly, minor forms of aggressive and delinq uent behavior precede engagement in more serious offenses, including substance use, which is typically followed by de sistance in adulthood. However, a small group of offenders may maintain high levels of aggressive and delinquent behavior well into adulth ood (Dodge et al., 2006). Historical trends in the pr evalence of more serious form s of aggression and delinquency during mid adolescence, including physical fighting and carrying a weapon, have been evaluated in the YRBSS biennially since 1991. Rates of other forms of aggr essive or deli nquent behaviors were not tracked by this survey. After st eady declines from 19912003, the prevalence of engagement in physical fights over the past year has since increased from 33% to 35.9% among adolescents. Decreases are also reported in other serious forms of aggression and delinquency from between 1991 and 1999 at which point no furt her change in rates has occurred. This includes carrying a weapon both in and out of school as well as carrying a gun during the previous month. These more recent trends highli ght the importance of continued research as well as evaluation and revision of current in tervention strategies in these areas.
24 Data from the YRBSS support previous resear ch findings that males engage in higher levels of aggressive and deli nquent behaviors than females (C DC, 2006). High school age males were more likely than females to have carried a weapon in the past month (29.8% vs. 7.1%) and to have been involved in a physic al fight (43.4% vs. 28.1%). Impor tantly, this gender difference is most notable for physically aggressive and delinquent behavior with some evidence that females may engage in higher or comparable leve ls of relational aggression (see Dodge et al., 2006, for a review of this literature). In addition there is some evidence that the gender gap in physical aggression may be narrowing, either gene rally, or in specific su bgroups of adolescents or adolescents living in certai n contexts such as urban communities (e.g., Nichols, Graber, Brooks-Gunn, & Botvin, 2006). Ethnic differences in ag gressive and delinquent behaviors reported in the YRBSS revealed that White students were less likely than Black an d Hispanic students to have been involved in a physical fight (33.1% vs. 43.1% and 41%), and to have been involved in dating violence (8.2% vs. 11.9% and 9.9%). There were equal reports of having carried a weapon in the past month (18.7%, 16.4%, and 19%) and of having carried a gun specifically (5.3%, 5%, and 6.5%) among White, Black, and Hispanic high school age adolescents. In summary, the historical trend in aggr essive and delinquent behavior has shown no changes in rates among high school students since 1999 with the ex ception of incr eases in the occurrence of physical fights since 2003. Preval ence of aggressive and delinquent behavior among high school students is high with 1 out of 3 adolescents havi ng been involved in a physical fight. Black and Hispanic males are at particularly high risk for engagement in aggressive and delinquent behavior Males in general, report ca rrying weapons more frequently
25 than females and engaging in physical fights more frequently; however, White males are less likely than Black or Hispanic males to have been involved in a physical fi ght or dating violence. Substance Use Initiation of substance use seems to fo llow a typical sequence beginning with experimentation with alcohol or tobacco in early adolescence (Newcomb & Bentler, 1989). Increases in use occur between ages 11 and 14. In itiation of marijuana use typically occurs after experimentation with tobacco or alcohol but precedes initiation of more serious substance use. Use of substances such as alcohol and tobacco do not necessarily lead to marijuana use and use of other substances such as coca ine, heroine, or other amphetamines and narcotics; however, use of such substances is almost always preceded by experimentation with alcohol and tobacco. This sequence in substance use initiation is describe d by the gateway drug model (Kandel, Kessler, & Margulies, 1978) which will be discussed in the following section. Between 1991 and 2003, the use of most substances among high school aged adolescents had been decreasing (CDC, 2006). However, current trends in the rates of substance use reveal no changes in use since 2003. This includes rates of lifetime (used substa nce one or more times during their life) as well as curr ent (used substance one or more times during the past 30 days) cigarette use, lifetime alcohol us e, current alcohol use, and ep isodic binge drinking, lifetime and current marijuana use, lifetime and current cocaine use, and lifetime inhalant use. This change in the trend highlights the continued importance of investigating mechanisms that underlie the initiation of substance use in adolescence. In addition, there are important differences in rates of substance use by both gender and ethnicity during the high school ye ars (CDC, 2006). For the most part, males show higher rates of drinking and driving (11.7% vs. 8.1%), cu rrent smokeless tobacco use (13.6% vs. 2.2%), current cigar use (19.2% vs. 8.7%), episodic binge drinking (27.5% vs. 23.5%), lifetime
26 marijuana use (40.9% vs. 35.9%), lifetime cocaine use (8.4% vs. 6.8%), lifetime illegal steroid use (4.8% vs. 3.2%), lifetime hallucinogenic drug use (10.2% vs. 6.8%), and lifetime ecstasy use (7.2% vs. 5.3%). Females and males have equal rates of current and frequent cigarette use (23.0% vs. 22.9% and 9.3% vs. 9.3%, respectively) lifetime and current alcohol use (74.8% vs. 73.8% and 42.8% vs. 43.8%, respectively), and li fetime methamphetamine use (6.0% vs. 6.3%). Females tend to have higher rates of lifetime inha lant use (13.5% vs. 11.3%) compared to males. Overall, Black adolescents tend to have lowe r rates of substance use compared to White and Hispanic adolescents, including current ci garette use (12.9% vs. 25.9% and 22%), smokeless tobacco use (1.7% vs. 10.2% and 5.1%), ciga r use (10.3% vs. 14.9% and 14.6%), lifetime alcohol use (69% vs. 75.3% and 79.4%) curren t alcohol use (31.2% vs. 46.4% and 46.8%), drinking and driving (4.9% vs. 11.3% and 10.5%), episodic binge drinking (11.1% vs. 29.9% and 25.3%), lifetime cocaine use (2.3% vs. 7.7% and 12.2%), lifetime inhalant use (6.8% vs. 13.4% and 13%), lifetime hallucinogen use (2.8% vs. 9. 4% and 9.4%), lifetime methamphetamine use (1.7% vs. 6.5% and 8.8%), and lifetime ecs tasy use (3.9% vs. 5.8% and 9.6%). White adolescents are more likely to engage in fre quent cigarette use (11.2% vs. 3.7% and 6.5%) and smokeless tobacco use, as well as episodic binge drinking compared to Black and Hispanic adolescents (see above). Hispanic adolescents are more likely than both Black and White adolescents to engage in lifetime cocaine, he roin (3.6% vs. 1.5% and 2.2%), methamphetamine, or ecstasy use. Rates of lifetime cigarette use (54.7%, 57.1%, and 54%) and current marijuana use (20.4%, 23%, and 20.3%) are equal among Blac k, Hispanic, and White adolescents. To summarize, the historical trend in substance use among high school students has not changed in recent years (since 2003) Prevalence of problem subs tance use is astounding with 1 out of 4 high school age adoles cents having engaged in binge drinking and roughly 1 out of 10
27 having driven after drinking. In addition, 1 out of 3 high school age adolescents have tried marijuana. White and Hispanic males appear to be at the highest risk for substance initiation. While there are no gender differen ces in cigarette and alcohol use, males do continue to have higher rates of binge drinking and drinking and driving as well as higher rates of more serious drug use. Black high school students engage in less substance use than their White and Hispanic peers. This highlights the need for further eval uation of the mechanisms leading to the initiation of substance use during early adol escence with attention to indivi dual differences or contextual variations. Moreover, understanding factors that promote or prevent the ons et of substance use as well as aggressive and delinquent behavior during adolescence is cr itical to promoting a healthy transition from adolescence into adulthood. Carrying a weapon, being involved in physical fights, current cigarette use, smokeless tobacco use, alcohol use, binge drinking, and marijuana use, as well as lifetime cocaine and inhalant use during adolescence contribute to the leading causes of death from adolescence to adulthood--motor vehicle cr ashes, homicide, and suicide during adolescence, as well as cardiovascular disease and cancer in adulthood (CDC, 2006). Given the relevance of preventing substan ce use, aggression, and delinquency during adolescence to the promotion of health in both adolescence and adulthood, numerous intervention strategies have been developed aime d at reducing the onset of these antisocial behaviors during adolescence. The first step to developing effective interventions involves understanding the etiology of substa nce use, aggression, and delinquency. Developmental Models of Aggression, Delinquency, and Substance Use Current literature provides several existing models for pathways for aggression and delinquency, as well as initiation of substance us e. There are a couple of general models of human development that are applicable to the deve lopment of antisocial behaviors. At the same
28 time, some models are specific to only one out come such as substance use (e.g., the gateway drug model). Other models, specific to the deve lopment of antisocial behaviors, can be applied to aggression, delinquency, and substance use; however, these models do not posit direct associations among these behaviors. The following section briefly reviews general developmental models, outcome specific models, ge neral models of antisocial behaviors, and models of associations among outcomes. The definition of broad terms such as deviant behavior or problem behavior is also addressed for relevant models. General Developmental Models A general developmental framework that has be en influential in understanding interactions between individuals and the context in which they live is Bronfenbre nners (1979) ecological model of human development. Th is model emphasizes the concept of an individual as embedded within, interacting with, influenc ing, and being influenced by a soci al context. For example, not only do contextual factors influence individual de velopment, but individu al characteristics also influence or form the social contexts in which i ndividuals interact. As discussed in the following sections, most models incorporate some elemen t of a person-context interaction while some more simplified models emphasize partic ular contexts or person factors. Social learning theory (Bandura, 1977) empha sizes the concept that individuals learn behaviors through the observati on and imitation of others as opposed to simply through punishment and reinforcement of behavior. He nce, observing substance use, aggression, and delinquency among adult role models, peers, an d even via media consumption can influence adolescent behavior. Moreover, when individuals see models rewarded or receive reinforcement for the behavior, the behavior is more likely to be modeled. Social learning theory emphasizes the role of multiple contexts on the development of antisocial behavior.
29 Outcome Specific Models Aggression and delinquency. Numerous individual and contextual factors have been determined to influence the development of aggression and delinquency. Notably, there is a great deal of similarity and overlap between predictors of aggre ssion and delinquency and predictors of substance use. In their most recent review, Dodge, Coie, and Lynam (2006) note that individual aggressive acts ar e largely determined by situationa l or contextual factors. In contrast, stable patterns of aggression are more often linked to individual diffe rence factors (e.g., genetics, temperament). However, aggressive behavior also develops contingently based on contextual factors (e.g., reinforcem ent of aggressive behavior) that may be stable over the course of development. Dodge and colleagues (2006) furthe r assert that the field is close to achieving a consensus model of the development of aggr ession based on the voluminous literature on childhood and adolescent aggression and the indivi dual and contextual factors that shape this development. In particular, a few theories have focused specifically on explaining the developmental patterning or pathways of aggressive or antisocial behaviors. Moffitts theory of adolescent limited versus life-course persistent deviant behavior posits two different life course trajectories for the deve lopment of deviant behavi or which includes both aggressive and delinquent behavior (Moffitt, 2006) Most individuals who engage in deviant behavior first begin during early adolescence, peak in mid-adoles cence, and subsequently decline upon entry into adulthood. This group is labe led adolescence limited offenders and research indicates that social/contextual factors are most salient to the development of deviant behavior among these individuals. A smaller group of individuals begin engaging in deviant behavior in childhood. This group evinces higher levels of deviant behavior throughout adolescence compared to the adolescence limited group and deviant behavior persists well into adulthood. This group is labeled life-course persistent o ffenders and research indicates that individual
30 characteristics are most salient to the developm ent of deviant behavior among these individuals. Moffitt and colleagues have presented a thorough analysis of gender differences and multiple risk factors associated with membership in th ese trajectories (2001). Additionally, other groups have documented similar characteristics and trajectories for an early-starter group that can be identified in the early school years and persists in aggressive behavior throughout childhood and adolescence (e.g., Broidy et al., 2003). Loeber and Hay (1997) outline three developmen tal pathways to boys antisocial behavior. The authority conflict pathway outlines the connection between oppositional and defiant behavior in childhood, such as bein g disobedient and disrespectful to adults, with aggressive and delinquent behavior in adolescen ce. The overt pathway begins with physical aggression which escalates over time into violen t behavior during adolescence. The covert pathway begins with minor forms of delinquent behavior (e.g., shoplifti ng) during early adolescen ce which escalate to more serious forms of delinquency (e.g., fraud, burgl ary) over time. This model emphasizes the risk associated with early initia tion of aggressive and defiant be havior on future engagement in violence and delinquency. Substance use. As is true for aggression and delinquency, a wide range of individual and contextual factors have been de termined to influence initiation of drug use with some factors proving to be drug specific while others have comparable effects across drugs. As noted, an influential behavioral model of substance use init iation is the gateway drug model (Kandel et al., 1978). The gateway drug model operates under the axiom that the best predictor of future behavior is past be havior. This model states that progres sion to more frequent substance use, substance abuse, and use of more serious substances (e.g. cocaine, heroin, etc.) is almost always preceded by experimentation with less serious substances (e.g. coffee, tea, alcohol, etc.). As
31 such, experimentation with the less serious substa nces serves as a gateway to more serious use but does not necessarily result in more seri ous use. The gateway drug model applies to substance use specifically and emphasizes the ro le of the individual in the progression of substance use above the role of context. Mayes and Suchman (2006) have recently proposed models illustrating pathways to the initiation of substance use as well as the transition from substance initiation to substance dependence and abuse. In both models, an emphasi s is placed on the role of genetic factors. Regarding substance initiation, the genetic ba sis for emotion regula tion and self-control contributes to later attitudes to ward social conventions and choi ce of peers. Those individuals with poor emotion regulation and se lf-control are at greater risk for substance initiation. This model also acknowledges the role of parental influence. The second model proposed by Mayes and Suchman (2006) illustrates the role of geneti c vulnerability in the transition from substance initiation to substance dependence and abuse. Genetic vulnerability towards dependence and mood disorders, in combination with substan ce initiation and engagement in antisocial behaviors, can lead to continued drug use, school failure and isolation, depression, and ultimately substance dependence and abuse. While these models do include engagement in antisocial behavior as a factor that is important in the transition from substance initiation to substance dependence, the main emphasis of these models is in illustrating the role of genetics on subsequent substance use. Models Applicable to Sev eral Antisocial Behaviors Several models have been proposed to expl ain the development of multiple antisocial behaviors including aggression, de linquency, and substance use. As indicated, whereas these models are often applied to aggression, delinquency, and substance use, they do not posit direct associations among these behaviors. Again, th ese models utilize aspects of more general
32 developmental theories like Bronfenbrenners ec ological model (1979) or social learning theory (Bandura, 1977). Similar to social learning th eory, the social development model (Catalano & Hawkins, 1996) asserts that antiso cial behavior is learned through interpersonal interactions. Antisocial behavior encompasses both delinquency and substance use in this model. The social development model differs from social learning theory in that it emphasizes bonding to antisocial others as an influence on antisoc ial behavior independent of social learning proce sses. Direct influences on future antisocial behavior include perceived rewards for antisocial behavior, belief in antisocial values, interaction with antisocial others, and attachment to antisocial others. The influence of past engagement in antisocial behavi or on future engagement in antisocial behavior is mediated via the aforemen tioned direct influences (C atalano & Hawkins, 1996). The social interactionist perspective emphasizes the role of the parent-child interaction in the development of antisocial behavior from childhood through adolescence (Patterson, DeBaryshe, & Ramsey, 1989). In this model, antisocial behavior encompasses a number of different behaviors including a ggression, delinquency, and substa nce use. This perspective follows a developmental progression towards more serious antisocial behavior which begins with maladaptive parenting practices. This leads to conduct problems and aggressive behavior in childhood which results in poor social skills, a cademic failure, and association with deviant peers. This in turn reinforces antisocial behavior, leading to mo re serious aggression, delinquency, and substance use. Botvin and Sheier (1997) pr esent a model which combines elements of the social interactionist perspective and a social cogniti ve perspective to illustrate the developmental processes underlying the emergence of violent an d drug using behaviors. Emphasis is placed on deficits in familial social interactions and soci al information processing. These deficits increase
33 the probability of associating with deviant peer s, aggression, delinquency, and substance use. While connections between substance use and aggr essive or delinquent behaviors are discussed, emphasis is placed on deficient social skills and lack of personal competence in the development of all three antisocial behaviors. The importance of intraindividual psychological characteristics such as sensation seeking and egocentrism are highlighted by the reckless be havior perspective (Arnett, 1992). Reckless behavior encompasses reckless au tomobile driving, sex without contraception, illegal drug use, and minor criminal activity. This perspective st ates that developmental changes in sensation seeking and egocentrism underlie the rapid increase in reckless behavior seen during the middle school years (ages 11 14), the relatively high le vels of reckless behavior maintained throughout adolescence, and the eventual decline observed in the 20s and early 30s. This perspective also emphasizes the importance of the interaction be tween psychological characteristics such as sensation seeking and the social environment in the expression of reckless behavior. The role of individual characteri stics is important because it al lows prevention scientists to develop targeted interventions for individuals at a particularly hi gh risk for antisocial behavior including initiation and escalation of substance use, aggression, and delinquency. The utility of targeting a personality characteristic such as sensation seeking directly as a method of intervention is a topic of debate Most researchers conceptualize sensation seeking as a stable trait which is unlikely to be reduced via intervention (Zuckerm an, 2007). However, it may be possible to change how sensation seeking is ex pressed, minimizing unhealthy expressions. It has been suggested that the expression of sensation s eeking is influenced via socialization and hence, is a potential target of intervention (Arnett, 1995). Despite the emphasis on person factors,
34 Arnett does acknowledge the importa nce of social contexts in the development of reckless behaviors such as substance use, aggression, and delinquency. Perspectives that focus on multiple domains of influence on antisocial behaviors are inherently complex, resulting in a large number of risk factors. Due to the large number of factors that influence initiation of substance use, as well as other antiso cial behaviors, some researchers have evaluated shear exposure to mu ltiple risk factors as a predictor of future antisocial behavior (Newcomb & Bentler, 1989). This includes an aggregate of behavioral, psychological, and social/contextual factors. As risk factors increase, the likelihood of initiation of substance use and engagement in other antisocial behaviors also increases. The previous models provide an excellent framework for understanding risk factors associated with substance use, aggression, and delinquency individually. This is particularly important during later childhood and early adolescen ce as this is the developmental time period during which onset of these behaviors are asso ciated with the great est risk of health compromising outcomes in later adolescence and adu lthood. Just as it is be neficial to evaluate risk factors from multiple domains, it is also beneficial to evaluate multiple antisocial behaviors as opposed to studying a single behavior in isol ation. By evaluating the co-occurrence of antisocial behaviors it is possible to determine common correlates among a va riety of outcomes. This is useful in the development of broad interventions aimed at reducing multiple antisocial behaviors as opposed to specific outcomes. In a ddition, studies of multiple antisocial behaviors allows for the evaluation of the impact that enga gement in one type of antisocial behavior may have on participation in other antisocial behavi ors. A great deal of information has been obtained about aggression, delinque ncy, and substance use from studies that focus on only one of
35 these outcomes; however, interrel ationships between antisocial be haviors can only be established via studies that evaluate multiple antisocial behaviors. Models of Associations between Outcomes To date, only two theories directly address the importance of evaluating multiple antisocial behaviors as opposed to individual outcome models. These are th e self-control theory adaptation of a general theory of crime (G ottfredson & Hirschi, 1990) and problem behavior theory (Jessor, 1992). A general theory of crime conceptualizes a ll antisocial behaviors (e.g., delinquency, aggression, substance use, etc.) as arising from a common underlying disposition towards general deviance. This would result in comm on correlates among all antisocial behaviors. Interventions aimed at addressing one such be havior should also impact other antisocial behaviors as well. This theory proposes that not only are all antisocial be haviors manifestations of the same underlying general deviance but that b ecause of this it is unne cessary to distinguish between different types of behaviors in the st udy of prevention and inte rvention. It is more important to identify the underlying common correla tes of these behaviors and focus intervention efforts on those correlates as this will result in the reduction of all antisocial behaviors. It is unnecessary, and in fact erroneous to try and develop interventi ons for aggression or delinquent behavior that are separate from in terventions for substance use. The general theory of crime pr oposes that all antisocial beha vior is a manifestation of a lack or loss of self-control or an inability to inhibit ones actions (Gottfre dson & Hirschi, 1990). Self-control is believed to be the underlying co mmonality associated with all forms of antisocial behavior. According to the general theory of cr ime, self-control is lear ned via parental training and socialization. Parental socialization strengthens bonds to conve ntional social institutions and this training is virtually comp lete by age 10. Hence, according to this theory, interventions
36 aimed at reducing antisocial beha vior should focus primarily on pa rental training to promote the development of socialization and self-control duri ng early childhood. Jessors problem behavior theory (1987) iden tifies psychological, so cial, and behavioral influences on antisocial behaviors. Problem behavior refers to all behaviors that depart from both the social and legal norms of society. Th is includes but is not limited to aggression, delinquency, and substance use, as well as precoc ious sexual intercourse. According to this perspective, problem behavior is similar to all ot her learned behaviors in that it is a functional and purposeful method of attaining ones goals. Hence, it generally increases upon entry into adolescence in response to many of the challenges that face adolescents. This includes but is not limited to attaining independence from parental authority, expressing opposition to the norms of conventional society, or st rengthening bonds with peers or youth culture. There is some evidence of ethnic differences in the applicability of Jessors problem behavior theory. Stanton and colleagues (1993) found that sexual activity belonged in a separate domain from other problem behaviors such as substance use and truancy among African American adolescents from resource depleted, urban environments. One explanation for this finding is that sexual behavior am ong this youth culture is not recognized as a problem behavior and hence, does not cluster predictably with other problem behaviors. Given that problem behaviors are defined as depart ures from cultural and societal norms, it is important to be cautious when conducting research on specific be haviors across cultures and subcultures. Problem behavior theory is similar to the ge neral theory of crime in that it postulates associations between various antiso cial behaviors. However, problem behavior theory is distinct from the general theory of crime in that it doe s not assert that it is unnecessary to evaluate specific antisocial behaviors. Rather, problem behavior theory emphasizes the importance of
37 identifying risk and protective factors for co-o ccurring antisocial beha viors. Jessor (1992) describes a person-situation inter actionist perspective that emphasi zes the role of both person and context in the development of antisocial behavi ors and points out limitations of focusing solely on one domain. Hence, problem behavior theo ry encourages evaluating multiple antisocial behaviors as well as interconnecti ons between them and risk factors to evaluate empirically the degree to which antisocial behavi ors are related to one another a nd stem from common causes. A strength of problem behavior theory and the general theory of crime is that both highlight the importance of commonalties between a variety of antisocial behaviors. However, these theories still have their limitations. By combining multiple antisocial behaviors into a construct of general deviance, information regarding potential influences of one antisocial behavior on another are lost. Identifying common correlates of multiple antisocial behaviors is indeed informative; however, predictive pathwa ys among antisocial behaviors are not specified by these frameworks. The Developmental Model of Reciprocal Influence The direction of effect re garding interconnections among antisocial behaviors has been evaluated differently across vari ous studies. In general, stud ies of late childhood and early adolescence evaluate the predic tive influence of aggressive or delinquent behaviors on later initiation and escalation of substance use and abuse (Farrell et al., 2005; Scheier & Botvin, 1996). In part, this is due to the fact that children rarely engage in substance use while aggressive behavior is not unusua l in childhood. In contrast, st udies of late adolescence through adulthood often evaluate the predictive influence of substance use on later delinquent or criminal behavior (Stacy & Newcomb, 1995). The question of the directionali ty or reciprocity of effects is an important issue to try and disentangle to further understand the development of these antisocial behaviors and esta blish age appropriate inte rvention techniques.
38 The Developmental Model of Reci procal Influence (DMRI) is m eant to serve as a heuristic device regarding the development of substance us e and abuse and other an tisocial behaviors (See Figure 1-1). Figure 1-1. The developmental mo del of reciprocal influence This model was informed by and is complementary to the aforementioned theories with the main distinction being an emphasis on unders tanding the interconnections between multiple antisocial behaviors. The basic framework of th e DMRI states that the interconnections between substance use and other antisocial behaviors, such as aggression or delinquency, are reciprocal when viewed broadly across the transition from childhood to adulthood. Fo r instance, there is Adult Aggression Adolescent Substance Initiation Social Influences Regular Substance Use and Abuse Child/Adolescent Aggression Psychological Influences Adult Criminal Behavior Adolescent Delinquency Biological Influences
39 evidence of aggression and delinquency predicti ng substance use and as well as evidence of substance use predicting aggression and delinque ncy (Farrell et al., 2005; Mason & Windle, 2002; Scheier & Botvin, 1996; Stacy & Newc omb, 1995; Vitaro, Brendgen, Ladouceur, & Tremblay, 2001). However, we propose that the specific direction of influence changes across the adolescent time period. Given that aggressive behavior is ofte n the first antisocial behavior to develop beginning in childhood (Dodge et al., 2006), aggressive behavior should predict the onset of delinquent behavior and substance use in early adolescence. In addition, given the higher levels of engagement in delinquent behavi or in early adolescenc e, delinquency should be predictive of initiation and escal ation of substance use in earl y adolescence. However, if substance use, in general, transitions to more fr equent use or abuse, the direction of association changes such that substance use is predictive of aggressive and delinquent behavior. This transition in the direction of effects will generally occur in mid to late adolescence with adolescent substance use becoming the strongest pr edictor of antisocial be havior in adulthood. According to this model, risk factors asso ciated with the development of aggression, delinquency, and substance use wi ll have the strongest influen ce during early adolescence. When evaluating adolescent predictors of antisocia l behavior in adulthoo d, adolescent substance use will have the strongest predictive value of adu lt antisocial behavior. Th e effects of other risk factors on adult antisocial behavior such as adolescent peer devi ance or sensation seeking, will be mediated by adolescent substance use. Hence, according to the DMRI, interventions aimed at reducing antisocial behaviors in general should focus on reducing risk factors associated with the development of aggressive and delinquent be haviors during childhood and early adolescence. This in turn will reduce the number of adolescents that initiate substance use and become regular substance users or abusers. Of course, there are some individuals who engage in
40 experimentation with substance us e in mid to late adolescence that do not have a history of aggressive or delinquent be havior (Dishion & Loeber, 1985; Loeber, 1988). While these individuals are at a lower risk of transitioning to substance dependence and abuse, they will still benefit from substance use interventions. Hence, by mid to late adolescence and early adulthood, interventions should more aggressively target reducing regu lar substance use or substance abuse. This strategy would reduce ad ult rates of other anti social behaviors. The DMRI is in line with other theories of antisocial behavior in emphasizing the importance of biological, psychological, and soci al factors in the development of aggression, delinquency, and the initiation of substance use during childhood and early adolescence. However, it differs from the prominent theories re viewed previously in two important respects. First, the DMRI asserts direct re ciprocal influences between speci fic antisocial behaviors rather than assuming that the co-occurrence of these behaviors is simply due to common antecedent risk factors. Second, the DMRI is not intended to be a general model of all aggressive or delinquent behavior. Rather, it focuses specifically on the association between substance use and aggressive or delinquent behavior and how this association develops from adolescence into adulthood. While only a handful of studies have eval uated reciprocal eff ects between antisocial behaviors, evidence to date lends support to the Developmental Model of R eciprocal Influence. This review summarizes results from studies which directly evaluate a predictive association between substance use and aggression or delinquenc y. Studies were excluded if (a) only a single antisocial behavior was evaluated, (b) multiple antisocial behaviors were combined into a construct of general deviance, or (c) co-occu rrence of multiple antisocial behaviors were
41 evaluated but predictive associations of one antisocial behavior on another were not. See Table 1-1 for methodological descriptio ns of the following studies. Table 1-1. Summary of studies that examined reciprocal effects among antisocial behaviors Authors Design Sample Measures Early Adolescence Margulies, Kessler, & Kandel, 1977 Longitudinal study of 18 public high school students (grades 9 12) in New York state (6 urban, 6 suburban, and 6 rural). Data were collected in 2 waves: beginning of academic school year (fall) and 5-6 months later. 1936 (62% female). 29% 9th graders, 31% 10th graders, 21% 11th graders, and 19% 12th graders. Race/ethnicity distribution was not provided. Self reported intrapsychic states (depression, self-image), academic orientation (classes cut, grade average, days absent, educational expectations), lifestyle values (conformity, political attitudes, church attendance), drug attitudes, delinquent involvement, prior drug use (beer, wine, and cigarettes), demographics (religion, race/ethnicity, family income, fathers education, gender). Parent reported substance use behaviors, attitudes towards substance use, parent/child relationship quality. Peer reported substance use, attitudes towards substance use, peer/adolescent relationship quality. Brook, Whiteman, Gordon, & Cohen, 1986 Longitudinal study of youth from one low SES county and one middle/high SES county in upstate New York. Data were collected over 2 waves: T1 ages 510, T2 ages 13-18. 356 youth (52% female), 94% Caucasian from diverse socioeconomic backgrounds. 51% were age 5-7 at T1 (age 13-15 at T2) Mother report of childhood personality and behavior (conventionality/unconventionality, control of emotions, intrapsychic functioning, and interpersonal relatedness). Mother and adolescent report of adolescent personality and behavior (conventionality/unconventionality, contol of emotions, intrapsychic functioning, and interpersonal relatedness). Adolescent self report of substance use.
42 Table 1-1. Continued. Authors Design Sample Measures Simcha-Fagan, Gersten, & Langner, 1986 Longitudinal study of Manhattan, New York children. Data is from a subsample followed up 5.5 years after baseline Baseline ages ranged from 8.5-18 years. Follow-up ages ranged from 14-23.5 years. 200 youth, 69.6% White, 12.9% Black, 17.5% Spanish speaking. Participants were a probability sample of Manhattan children. Gender distribution and SES were not provided. Mean age at Follow-up assessment 18.3 years. Mother reported social demographic characteristics (mothers education, monthly rent), familial dimensions (parent relationship quality, relationship quality with child, parenting behaviors, mothers physical and mental health), and child behavior (parent-child conflict, fighting, delinquency, non-compulsive antisocial tendencies, isolation from peers, anxiety, self-destructive tendencies, dependence). Adolescent/young adult self reported legal and illicit substance use (alcohol, cigarettes, marijuana, lsd, methedrine, amphetamines, barbiturates, heroin) Scheier & Botvin, 1996 Cross-sectional study of 8th graders from a low SES, urban area in New York 418 African American 8th graders (51% female) from low SES Self reported measures made up 4 latent constructs; general deviance (unconvential behavior, sensation seeking, aggressive behavior, risk taking behaviors), polydrug use (frequency of cigarette smoking, alcohol use, marijuana use, and inhalant use as well as intensity of alcohol and marijuana use), personal anomie (loneliness, future hope and life purpose, hopelessness, existential purpose in life and meaning, and suicidal ideation), and cognitive efficacy (personal competence and cognitive mastery, self-reinforcement, applied decision-making skills, and self-esteem). Farrell, Sullivan, Esposito, Meyer, & Valois, 2005 Longitudinal study of youth from 3 urban middle schools and 4 rural middle schools in a low SES area in south eastern United States. Data were collected over 5 waves: the beginning and end of 6th and 7th grade and the beginning of 8th grade. 667 urban adolescents (51% female), 96% African American, 3% Caucasian, mean age at first assessment 11.7 years; 950 rural adolescents (47% female), 58% Caucasian, 24% Latino, 14% African American. Self reported aggression, non-violent delinquency, drug use (average of cigarette use, alcohol use, and marijuana use)
43 Table 1-1. Continued. Authors Design Sample Measures Lillehoj, Trudeau, Spoth, & Madon, 2005 Longitudinal study of 7th grade students drawn from 36 rural schools in the mid western United States. Data were collected over 5 waves: baseline, 6, 18, 30, and 42 months after baseline. 198 youth (49% female), 96% Caucasian from middle class background, mean age at baseline assessment 12.3 years Self reported antisocial behaviors (CBCLYSR, disobediance, misconduct, aggression), number of substances initiated (cigarettes, alcohol, marijuana) Late Adolescence and the Transition to Adulthood Stacy & Newcomb, 1995 Longitudinal study of individuals from the Southern California area. Data were collected in 2 waves: late adolescence and early adulthood. 536 predominately Caucasian, middle class individuals (72% female), mean age 18.9 years at first assessment and 26.9 years at follow-up Self reported general drug use (cigarettes, alcohol, marijuana, hard drugs, and cocaine), peer deviance, socialemotional security, social support, academic orientation and family disruption, social conformity, and criminal deviance Ensminger & Juon, 1998 Longitudinal study of youth drawn from the urban, low SES Woodlawn neighborhood located in the south side of Chicago. Data were collected in 4 waves: 2 early childhood, 1 adolescence, and 1 adult. 953 predominantly African American youth (52% female) from an urban, low SES background. Assessed at 1st grade, 3rd grade, 10th grade, and at age 32-33 Teacher reported school behavior and grades in 1st grade. Self reported adolescent antisocial behaviors (physical assault, marijuana and hard liquor use), adolescent social bonds (parental supervision, attachment to school, attachment to mother), transitions to adulthood (marriage, childbearing), adult antisocial behaviors (alcohol, marijuana, cocaine use, interpersonal aggression), adult SES, and anxiety
44 Table 1-1. Continued. Authors Design Sample Measures Reciprocal Influences Vitaro, Brendgen, Ladouceur, & Tremblay, 2001 Longitudinal study of males from disadvantaged neighborhoods in Montreal, Quebec, Canada. Data were collected in 2 waves: early adolescence and late adolescence. 717 French-speaking Caucasian males, age 13 14 at first assessment and age 16 17 at follow-up Self reported impulsivity, friends' deviancy (illegal activity), parental supervision, and sociodemographi c info (family configuration, parental occupation) at age 13 14. Self reported gambling frequency, gambling problems, delinquency, and drug/alcohol problems at age 16 17. Mason & Windle, 2002 Longitudinal study of youth drawn from 3 suburban highschools in western New York State. Data were collected in 4 waves at 6 month intervals. 1218 predominantly Caucasian middle class youth (51% female), mean age at first assessment 15.51 years Self reported aggression, property damage, theft, cigarette use, marijuana use, and alcohol use in the past 6 months Early Adolescence Looking at the first part of the model, a few studies have specifically examined the predictive role of aggression or delinquency on substance use initi ation in early adolescence. Margulies, Kessler, and Kandel (19 77) evaluated factors associated with the transition to hard liquor consumption among high sc hool students (grades 9 12). This study only evaluated students from the original sample who had previously never consumed hard liquor which constituted only 35% of the original sample. Hence, these results may not generalize to the majority of adolescents who initiated hard liquor use earlier. Neve rtheless, overall results indicated that engagement in minor delinquent ac tivity was one of the be st predictors of the transition to hard liquor consumption. When evaluated separately by gender, minor delinquent activity was predictive of hard liquor consum ption among males only. However, peer and
45 parental alcohol use exert nearly three times the influence on female use of hard liquor compared to males. Hence, among youth who tend to be more resistant to substance use, minor delinquency predicts hard liquor use for male s while females may be more influenced by interpersonal role models. Brook, Whiteman, Gordon, and Cohen (1986) lo ngitudinally assessed both childhood and adolescent personality and behavi oral factors associated with adolescent substance use in a predominantly White sample of males and females. Results indicated that mother report of childhood delinquency was associated with adolescen t delinquency which in turn predicted stage of substance use in adolescence. The effect of childhood delinquency on stage of adolescent drug use was mediated by adolescent delinque ncy. This study did not evaluate gender differences or differences be tween older and younger adolescen ts. These variables were controlled for along with socioeconomic status. In a similar study, Simcha-Fagan, Gerston, a nd Langner (1986) longitudinally evaluated the association between mother report of child/adolescent behavior a nd adolescent/young adult self report of illicit substance use five and a half y ears later. Participants were divided into four groups of illicit substance use: no use, ma rijuana only, substances other than marijuana (excluding heroin), and heroin use. When cont rolling for social demographic characteristics and familial dimensions, early childhood/adolescent delinquency surfaced as a unique predictor of the transition to marijuana use as well as th e transition to heroin use in adolescence/young adulthood. Similarly, early chil dhood/adolescent aggressive beha vior, such as fighting, was predictive of the transition to substance use othe r than marijuana (excludi ng heroin) controlling for familial dimensions. While gender and race/ ethnicity differences were not evaluated for pathways between antisocial behaviors, differences in stage of illicit substance use were
46 evaluated. White adolescents/young adults were more likely to have transitioned to marijuana use as well as use of illicit subs tances other than marijuana. This is congruent with current national trends as discussed previously (CDC, 2006). Males were significantly more likely to have used heroin. No other gender or ra ce/ethnicity differenc es were observed. In a cross-sectional sample of Black middle school students, a latent construct of general deviance mediated the relationship between both personal anomie and cognitive efficacy on polydrug use. In this study, general deviance included both behavi ors, such as aggression and risk taking, as well as personality factors, su ch as sensation seeking and unconventionality (Scheier & Botvin, 1996). There were no ge nder differences on any of the drug use items; however, males reported higher rates of physical aggression, sensation seeking, and risk taking behaviors while females reported higher grades, conventionality, self-reinforcement, self-esteem, and decision making skills. Levels of drug us e were somewhat lower than in national and regional samples. This is cons istent with current trends indi cating that Black youth in general have lower rates of drug use than White or Hispanic youth as di scussed previously. A longitudinal study of the same early adoles cent time period evaluated quadratic growth curves of aggression, delinquency, and substance use (F arrell et al., 2005). Ov erall, initial levels of aggression predicted cha nge in drug use and delinquency. While females had lower initial rates of drug use, aggression, and delinquency in this sample, there were no differences by gender in pattern or rates of change for any of the antisocial behaviors across middle school. This study did not evaluate ethnic differences but did evaluate differences between urban and rural environments. This was largely confounded with ethnicity such that 96% of adolescents described themselves as Black in the urban sa mple while the rural sample was 58% White, 24% Hispanic/Latino, and 14% Black. Urban adolescent s had higher initial levels of aggression and
47 delinquency and greater increases in aggression, delinquency, a nd drug use compared to rural adolescents. Importantly, a general antisocial beha vior factor did not acco unt for the data as well as the models evaluating direct effects between separate antisocial behaviors. Gender differences in associations between s ubstance initiation and antisocial behaviors (aggression, disobedience, and mi sconduct) were evaluated long itudinally among a sample of rural White adolescents (Lillehoj, Trudeau, Spoth, & Madon, 2005). There were no gender differences in antisocial behaviors in the 7th grade; however, males did report higher levels of substance initiation in the 7th grade compared to females. Among both genders, 7th grade antisocial behaviors predicted in itiation of substance use but not change in substance use over time. This provides evidence in support of a link between aggre ssion and initiation of substance use. Notably, longitudinal studies of childhood aggr ession and subsequent adolescent behavior are nearly absent from the literature. Due to costs and time constraints, most studies of adolescent substance initiation ha ve focused on concurrent behavior s or longitudinal associations within the adolescent de cade or period of adolescence (e.g., mi ddle school). Interestingly, there are projects currently collecting and analyzing data that will be able to address this gap in the literature. For example, the Conduct Problem s Research Prevention Group (e.g., 1992, 2004) is conducting a multi-site evaluation of a long-term intervention program that presently has complete cohort data available for kindergarten through 5th grade. In addition, the Oregon Youth Substance Use Project (OYSUP; Andrews, Tildesley, Hops, Duncan, & Severson, 2003; Andrews, Hampson, Barckley, Ge rrard, & Gibbons, in press; Ha mpson, Andrews, Barckley, & Severson, 2006) is a cohort-sequential evaluation of individual a nd contextual influences on the acquisition and development of substance us e from early childhood (kindergarten) through
48 adolescence (12th grade). On-going research from thes e projects is well positioned to examine pathways to substance use from childhood through adolescence. Late Adolescence and the Transition to Adulthood Turning to the later part of the model, ag ain, only a few studies have examined prediction from adolescence into adulthood. In a predominantly White sample, Stacy and Newcomb (1995) evaluated the influence of late adolescent drug use on adult crimin al deviance, controlling for the possible predictive effects of other adolescent f actors including peer deviance. Adolescent drug use was found to be the only significant path predicting criminal deviance in adulthood. However, evaluations of specific drugs did not predict criminal deviance in adulthood. Unfortunately, gender effects were not evaluate d due to a highly skewed gender distribution (72% female). Hence, gender was controlled for in relevant analyses. Despite this limitation, these findings lend some support to the DMRI in that a general factor of polydrug use in late adolescence predicted criminal deviance in ad ulthood, unmediated by differential association with deviant peers, specific drug effects, or attitude. Ensminger and Juon (1998) evaluated pattern s of antisocial behaviors in adulthood among the participants of the Woodlawn longitudinal stu dy of children. This st udy evaluated a sample of African American children from an impoverish ed community on the South Side of Chicago. For both males and females, heavy use of liquor and marijuana during mid adolescence (approximately 10th grade, ages 15 16) predicted membership in clusters of adulthood antisocial behaviors characterized by high substance use and interp ersonal aggression or violent behavior assessed at ages 32 33. While these high risk indivi duals only made up 20.2% of the males and 21.7% of the females in the sample, th ey represent the group w ith the most risk for whom a targeted intervention program is most necessary. Among individuals who did not engage in high rates of substance use in adolescence, there were markedly lower rates of
49 problems in adulthood. Other factors from early childhood and adolescence also differentiated the clusters. Unfortunately, the influences of the childhood and adolescent factors on adulthood cluster group membership were evaluated independe ntly of one another. Hence, the relative influence of any one factor compared to the others was not determined. Reciprocal Influences In contrast to these uni-di rectional studies of influence, Vitaro, Brendgen, Ladouceur, and Tremblay (2001) utilized structural equation m odeling (SEM) to test predictive relationships between gambling, delinquency, and a general drug and alcohol use construct in a sample of White males in Canada. Only drug and alcohol use at age 16 was pred ictive of the other antisocial behaviors at age 17. Impulsivity, frien d deviancy, and parental supervision in early adolescence (ages 13 and 14) expl ained a significant, although small, amount of the residual covariance between the outcomes illustrating that covariance between drug and alcohol use, delinquency, and gambling in mid adolescence ar e partially accounted for by early adolescent predictors. Unfortunately, this study could not ev aluate the moderating effects of gender due to the absence females in the sample. However, these findings highlight the importance of understanding adolescents who engage in multiple antisocial behaviors ra ther than just one antisocial behavior as the etio logy and prognosis may differ be tween these groups. This study provides important information re garding reciprocal influences between antisocial behaviors given that substance use in mid adolescence was predictive of other antis ocial behaviors but the reverse association was not substantiated. Only one other study has examined directiona l associations between antisocial behaviors and substance use. Mason and Windle (2002) l ongitudinally evaluated r eciprocal influences between substance use and delinquency in a sa mple of White males (49%) and females. Participants were 15.5 years of age at firs t assessment and were reassessed at 6th month intervals
50 for 2 years. Results indicate consistent but small associations between delinquency at an earlier time point and substance use at the subsequent time point among males only. Substance use at first assessment was also positively associated with delinquency at second assessment among males. Evidence of reciprocal influences was not found for females. These results also lend support to the DMRI; however, this study is limite d in that it encompasses a very small window of time in adolescent development and it does not address the early adolescent time period, during which the initiation of these antisocial behaviors actually occurs. Salient Issues for Understandi ng Reciprocal Influences Most research that evaluates antisocial behavi ors either examines each behavior separately as individual outcomes or antisoci al behaviors are combined into a construct of general deviance a priori. Results of these st udies do provide meaningful info rmation regarding common risk factors for antisocial behaviors but ignore the pos sibility that engageme nt in one antisocial behavior may serve as a risk for engagement in ot her forms of antisocial behaviors. At the same time, as can be seen from the brief list of publ ished studies reviewed, examination of reciprocal influences has been lacking. In addition, studies to date have not fully examined issues of diversity in these developmenta l processes. It should be not ed that one drawback to the examination of reciprocal influences has been an absence, until recently, of appropriate analytic methods for testing the DMRI and other developmental models. Gender. Few studies have examined the role of ge nder in regards to reciprocal influences between substance use, aggre ssion, and delinquency. Of these, reports of engagement in aggression, delinquency, and substance use either show no differences by gender or males report higher levels of engagement. As noted, Mas on and Windle (2002) found re ciprocal influences between delinquency and substance use only for ma les. However, evaluations across a longer range of development (from early adolescence to early adulthood) may result in different
51 patterns of associations between antisocial behaviors. This encourages further exploration of gender differences when evaluati ng reciprocal influences. Ethnicity. Moreover, ethnic diversity is woefully underrepresented in evaluations of reciprocal influences among antis ocial behaviors. As discusse d previously, there is some evidence that certain behaviors wh ich are generally seen as problematic (i.e., precocious sexual activity) do not necessarily cluster with other antisocial behaviors among African American adolescents (Stanton et al., 1993). Whereas, Ensminger and Juon (1998) found that heavy substance use in late adolescence was associated with adulthood aggression and delinquency in a sample African American males and females. Of course, the W oodlawn study was conducted with only urban, low-income children. Given ethnic differences in prevalence of substance use, indicating lower use during ear ly to mid adolescence among Af rican American youth, and aggression and delinquency, indicating higher rates for this group in comparison to White youth, it would be informative for future research of reci procal influences to ev aluate ethnically diverse samples within a variety of regional and socio-ec onomic contexts. At present, it is not clear whether ethnicity is particularly informative in understanding pathways fo r antisocial behaviors or whether confounding contextual factors (e.g., urban versus s uburban environments) or socioeconomic factors (e.g., poverty, pare ntal education, employment) are more salient for prediction. Clearly, these distinctions are important for prevention programming and accurately identifying not only who may be at heightened risk but also who may be protected from risk, and most importantly, why risk or protecti on is conferred (Newcomb, 1995). Analytical approaches. As discussed previously, nu merous studies have examined individual differences in beha viors and changes in behaviors across time via comparisons of time-specific mean levels of a particular outc ome (e.g., aggression) for a particular group (e.g.,
52 males). While these studies do provide informati on that can be quite useful, they are limited when attempting to understand outcomes which ar e inherently developing over time (Curran & Muthen, 1999). Frequently, it is of interest to understand how specific factors alter the normative developmental trajectory of a particular outcome across multiple time points rather than simply evaluating change in that outcome between two discrete time points. This section addresses three different analytic techniques used to evaluate longitudinal data: Hierarchical Linear Modeling, Structural Equation Modeli ng, and Group-based trajectory analysis. Hierarchical Linear Modeling (HLM) allows researchers to evaluate changes that occur over time within an individual as well as differences between i ndividuals (Tabachnick & Fidell, 2007). This allows researchers to account for the nested nature of assessing the same individuals across multiple time points by allowing them to model intraindividual change as well as interindividual variability over time This has been useful in th e study of substance use and other antisocial behaviors becaus e researchers are able to statistica lly evaluate changes in individual substance use over time as well as potential differences in changes between genders or ethnicities. This method is very similar to re gression and hence, has the same advantages and limitations. Moderating effects are frequently evaluated using this technique; however, structural equation modeling is often recommended to eval uate mediating influences. Structural Equation Modeling (SEM) allows researchers to stat istically evaluate associations among both observed (measured) variables and latent vari ables (Tabachnick & Fidell, 2007). This analytic technique provi des the clearest assessment of both direct and indirect (mediated) effects. In fact, four of the studies described in Table 1-1 utilize this approach in evaluating associ ations among antisocial behavior s (Mason & Windle, 2002; Scheier & Botvin, 1996; Stacy & Newcomb 1995; Vitaro et al., 2001). One application of SEM, latent
53 growth curve modeling (LGC), is conceptually very similar to HLM. LGC allows researchers to evaluate individual change in one or more domains over time (Tabachnick & Fidell, 2007). It is similar to HLM in that one can assess intraindiv idual change over time as well as interindividual variability. Of the studies repor ted in Table 1-1, Farrell et al. (2005) and Lillehoj et al. (2005) both employ linear growth curve modeling to ev aluate longitudinal a ssociations between antisocial behaviors. LGC does have a few a dvantages over HLM under certain circumstances. Both HLM and LGC can be used to calculate para llel growth curves of more than one outcome; however, LGC is generally easier to interpret an d is more frequently used for this type of analysis. In addition, LGC is capable of estimating time intervals as well as using growth to predict future outcomes, neither of which is possible with HLM. Group-based trajectory analysis assesses developmental traject ories for group-based rather than individual growth curves (Nagin, 1999; Nagin & Tr emblay, 2001). This method is used to identify distinctive groups of individual trajectories within samples with multi-wave data. This method includes the capability to: (a) relate group membersh ip probability to individual characteristics and circumstances; (b) use the gro up membership probabilities to create profiles of group members; (c) add time-vary ing covariates to trajectory models; and (d) estimate joint trajectory models of distinct but related behavi ors. The advantage of group-based trajectory modeling is the ability to identify distinct groups of trajectories within the population. This is in contrast to HLM and LGC modeling, both of which assume a conti nuous distribution of trajectories within the p opulation. Assuming a con tinuous distribution of tr ajectories within the population is in direct opposition to the idea of distinct clusters or categories of development and therefore it is awkward to use these methods wh en evaluating research questions that address developmental trajectories that are inherently categorical.
54 Notably, a debate regarding the application and interpretation of group-based trajectory modeling, as well as other complex longitudinal statistical methods, has arisen in response to the recent increase in the use of these analytic te chniques. Numerous articles have centered around this debate1 (Nagin & Tremblay, 2005a; Nagin & Tr emblay, 2005b; Sampson & Laub, 2005). In brief, caution must be used when applying these complex methodologies so as not to misinterpret the results obtained. For example, regarding gro up-based trajectory analysis it is important to emphasize that individuals do not actually belong to trajectory groups, the number of groups is not immutable, and even individuals with a high probability of belonging to a particular group do not follow that group trajectory exactly (Nagin & Tremblay, 2005a). Given these features of group-based trajectory modeling, it has been argued that result s provided by this type of technique are potentially flawed due to ambigu ity in groups and group membership, an inability to accurately predict individual outcomes based on group membership, and a tendency to rely on results of statistical analyses to draw conclusions and inform future analyses as opposed to relying strictly on theory (Sampson & Laub, 2005) Despite these criticisms, utilizing new statistical methodologies can provide important new information that can then be useful in the revision and refinement of existi ng theories as well as the development of new theories, provided the analyses are applied and interpreted appropriately. Conclusions The empirical evaluation of the development and prevention of antisocial behaviors is a challenging endeavor. Our current state of knowledge has benef ited immensely from numerous, large, nationally representative longitudinal studies of antisoc ial behaviors. Longitudinal research designs are essential in disentangling the reciprocal influences between antisocial behaviors from childhood to adulthood. In addition, advances in statistical methodology, including trajectory analysis a nd structural equation modeling, allow researchers to evaluate
55 more accurately the associations among construc ts over time. Finally, continued development and evaluation of developmentally appropria te intervention techniques, that includes interconnections between antisoc ial behaviors across the lifespan, are critical in making the translation from empirical research to real world applications. Measurement issues associated with accura tely assessing initiation of and continued engagement in antisocial behaviors will alwa ys be a challenge. Self report remains an informative method of assessing antisocial beha viors. Analyzing saliva samples for carbon monoxide and thiocyanate levels has not been esta blished as a valid alte rnative to self-reported tobacco use among adolescent populations (Snow Gilchrist, & Schinke, 1985). However, collecting saliva and breath samples can enhance the validity of self report responses (e.g., bogus pipeline procedure), as does emphasizing confidentiality of responses (Evans, Hansen, & Mittelmark, 1977). Co-occurrence of antisocial behaviors is quite common throughout adolescence and adulthood. Unfortunately, literat ure on the development of antis ocial behaviors is currently limited by evaluating each antisocial behavior se parately or by combining them into general deviance constructs without t horoughly evaluating reciprocal influences among antisocial behaviors. Clearly there are some individuals who engage in only one form of antisocial behavior (e.g., delinquent or crim inal behavior); however, the a ssociation between substance use and delinquent or aggressive behavior is hi ghlighted by one study of youth with pervasive antisocial behavior. This st udy found that 51% of youth with pervasive antisocial behavior reported high levels of substance use compared to 11% of youth without pervasive antisocial behavior (Tiet, Wasserman, Loeber, McReynol ds, & Miller, 2001). Adolescents who are engaging in multiple antisocial behaviors repres ent a significant subgroup of individuals at
56 higher risk for continued problems in adult hood. Having an accurate understanding of the reciprocal effects of one antisocial behavior on another across adolescence will be invaluable in the development and implementation of appropriate intervention strategies for these high risk youth. The Developmental Model of Reciprocal In fluence (DMRI) attempts to highlight developmental differences across adolescence rega rding interconnections between substance use and other antisocial behaviors. It emphasizes the multifaceted pathways to the development of individual antisocial behavior s in early adolescence while re cognizing the interconnections between antisocial behaviors in late adolescen ce and adulthood. As recommended by Newcomb and Bentler (1989), DMRI distinguishes between substance initiation and experimentation as opposed to regular substance use and abuse. Regular substance use and abus e is associated with problematic adult outcomes and criminal beha vior. Importantly, DMRI also recognizes the importance of reducing the initiation of antisocial behaviors in early adolescence, as escalation of antisocial behaviors begi ns with initiation. The DMRI has strong implications for interventions involving substance abusers who also engage in other antisocial beha viors. Preventing the onset and escalation of antisocial behaviors in adolescence has long been a prio rity of policy makers, prevention scientists, and the public in general. Initial strategies focusing on subs tance use prevention, util ized dissemination of information about the harms of substance use as well as the Just Say No campaign (Lynam et al., 1999). These first forays into substance prevention proved to be largely unsuccessful. Hence, more rigorous scientific approaches to the development and evaluation of substance use prevention interventions have burgeoned (e.g., Life Skills Training Program, Botvin & Griffin, 2004).
57 Clearly, current interventions/programs aime d at reducing risk factors and enhancing protective factors during childhood and early adolescence are imperative. However, programs aimed at early identification of regular substance users or abusers who engage in other antisocial behaviors should strongly emphasize substance use rehabilitation as a mechanism through which future adult criminal activity and ag gressive behavior will be reduced. The DMRI supports the evaluati on of biological, psychological, and social risk factors on the onset of antisocial behaviors in early adolescence. However, further research is needed to evaluate if these risk factors operate directly or indirectly through regular substance use and abuse during late adolescence to predict future adult criminal behavior and aggression. Importantly, the DMRI should be evaluated am ong both males and females from a variety of ethnic backgrounds, regional cont exts, and socio-economic strata given that there is evidence of gender and ethnic differences in prevalence of antisocial behaviors, as well as differential effects of risk and protective factors (Andrews, 2005; Eide, Acuda, Khan, Aaroe, & Loeb, 1997; Mason & Windle, 2002; Newcomb, 1997; Svenss on, 2003). Information regarding gender and ethnic differences in reciprocal influences betwee n antisocial behaviors is sparse and inconsistent thus far, requiring further examination. Longitu dinal evaluations beginning in early adolescence, before many antisocial behavior s are initiated, and continuing through the transition to adulthood will provide the most complete picture of associ ations between antisocial behaviors as well as the influence of risk and protective factors. The Proposed Study The DMRI is meant to serve as a heuristic de vice to guide future research. Given the breadth of the model, it would be unrealistic to a ttempt to evaluate the model in its entirety. As such, the proposed study begins to address some of the gaps in the literature regarding the DMRI by focusing on one particular part of the model. Specifically, associations delineated in the first
58 part of the DMRI, relating traject ories of aggressive behavior to trajectories of delinquent behavior as well as both of th ese with initiation of substance use during early adolescence. These pathways will be evaluated in a longitudinal sample of ethnically diverse males and females.
59 CHAPTER 2 THE ROLE OF SENSATION SEEKING AND DEVIANT PEER ASSOCIATION ON ANTISOCIAL BEHAVIOR The middle school years (ages 11-14) are when behaviors such as substance use (tobacco, alcohol, marijuana, and other dr ugs), aggression, and delinquency increase more dramatically. While a great deal of information has been gain ed by evaluating these co nstructs independently of one another, it is important to investigat e the interconnections betw een the development of aggressive and delinquent behaviors and the development of substance use and abuse in adolescence and adulthood (Jessor, 1992). Resear ch has shown that rates of all three of these behaviors increase during this time frame and th at they share common co rrelates, yet it is still unknown whether aggression and delinquency deve lop temporally along with drug use or whether one naturally precedes or leads to the ot her. Equally importantly, does this association vary by gender? Previous resear ch has investigated both sensati on seeking, and association with delinquent or drug using peers with the development of aggre ssive, delinquent, and drug using behaviors in adolescence. However, the asso ciation between these factors and the temporal relationship between aggression, delinquency, and drug use in mi ddle school has yet to be established. Pathways to Drug Use and Aggression and Delinquency: Conceptual Framework A wide range of individual and contextual factors have been determined to influence initiation of drug use during the middle school ye ars with some factors proving to be drug specific while others have comparable effects across drugs. Current lit erature provides several existing models for initiation of drug use (B otvin & Sheier, 1997; Catalano & Hawkins, 1996; Chassin & Ritter, 2001; Newcomb & Bentler, 1 989; Sher, 1991) as well as pathways for aggression and delinquency (Bandura, 1977; Broi dy et al., 2003; Crick & Dodge, 1994; Jessor, 1992; Moffitt et al., 2001). This project draws specifically from Jessors problem behavior
60 theory (Jessor, 1992). An advantage of Jessors th eory (1992) is the broa d conceptualization of problem behaviors rather than a focus on speci fic outcomes. Jessors theory (1992) also emphasizes the importance of identifying risk an d protective factors for co-occurring problem behaviors. I am extending this model by incorporating new and innovative methods for examining co-occurrence of problem behaviors that allows for examining temporal associations and pathways, with attention to individual and peer factors that set youth on different pathways/trajectories. My framework for studying adolescent drug use draws heavily from Bronfenbrenners (1979) ecological model of human development. This model emphasizes the concept of an individual as embedded within, interacting with, influencing, a nd being influenced by a social context. For example, not only do contextual factors influence indivi dual development, but individual characteristics also influence or form the social contexts in which individuals interact. It is clear that the ecological model as well as other models drawn upon for the proposed study include multiple factors from several contexts (cultural/societal environment, interpersonal forces, psychobehavioral factors, and biogenetic influences). Rath er than attempting to evaluate this complex model in its entirety, my goal is to provide a better delineation of pieces of the model that may be particularly salient to the interconne ction of pathways to drug use, aggression, and delinquency. For this project, the primary focus is on the hypothetical model shown in Figure 3-1. Specifically, I am interested in evaluating ge nder-specific and common pathways for aggression and delinquency and the interconnection to pathways to drug use. As discussed in the previous chapter, it is important to evaluate the role of individual characteristics to aid in the development of targeted interventions for i ndividuals at risk for the develo pment of antisocial behaviors.
61 While most researchers conceptualize sensation seek ing as a stable trait, unlikely to be reduced via intervention (Zuckerman, 2007), Arnett (1995) argues that it ma y be possible to change how sensation seeking is expressed, minimizing unhealthy expressions. In addition, Arnett acknowledges the importance of soci al contexts in the developmen t of substance use, aggression, and delinquency. The formation of healthy peer relationships con tinues to be a key element of current drug use and violence prevention pr ogramming (e.g., Dishion & Loeber, 1985; Kandel & Andrews, 1987). As such, it is important to unders tand the role of sensa tion seeking and deviant peer association regarding changes in antisoc ial behavior from chil dhood through adulthood to better inform the aforementioned intervention strategies. Figure 3-1. Hypothetical mode l of individual and contextu al pathways to drug use Individual sensation seeking as we ll as peer contexts will be investigated as predictors of pathways for aggression and delinquency and the interconnection to pathways to drug use. The evaluation of sensation seeking and peer context in connection to studying pathways is unique in that previous studies have evaluated these construc ts separately for each outcome, rather than as predictors of pathways between drug use, aggression, and delinquency. It is likely that the interrelationship among these f actors varies by gender; hence, gender is evaluated as a moderator. The proposed study draws from and ela borates upon models that have served as the Psychological Influences: Sensation Seeking Social Influences: Peer Devianc y Adolescent Aggression and Delinquency Adolescent Drug Use *Note: Moderating effects of gender and race/ethnicity will be evaluated for this model
62 basis of previously developed violence and substance use prevention programming (Botvin & Sheier, 1997; Catalano & Hawkins, 1996). Therefore, the results of the present project will directly inform future prevention programming efforts. Core Domains Individual Sensation Seeking Measurement considerations. Sensation seeking is the need for varied, novel, and complex sensations and experiences and the willingness to take phys ical and social risks for the sake of such experiences (Zuckerman, 1979). Studies that have evalua ted the construct of sensation seeking have found it to be a fairly st able biologically based personality trait (Bardo & Mueller, 1991; Clayton, Leukefeld, Donohew, Bardo, & Harrington, 1995; Bardo, Donohew, & Harrington, 1996). Sensation seeking is traditionally conceptualized as being composed of four subscales: experience seeki ng, thrill and adventure seek ing, disinhibiti on, and boredom susceptibility (Zuckerman, 1986). Zuckerman (1986) developed the original self report questionnaire used to evaluate this construct. Reduced forms of the original questionnaire have been developed and widely used, including a four item version where each item represents one of the four subscales. Research that has evaluated the effects of sp ecific sensation seeking subscales on antisocial behaviors have found stronger associations betw een antisocial behaviors and the thrill and adventure seeking and disinhibi tion subscales (Alexander, A llen, Brooks, Cole, & Campbell, 2004; Andrucci, Archer, Panc oast, & Gordon, 1989; Bardo et al., 1996; Donohew, Lorch, & Palmgreen, 1991; Donohew et al., 1999; Kopstei n, Crum, Celentano, & Martin, 2001; Newcomb & McGee, 1991; Zuckerman, et al., 1990). Thrill and adventure seeking refers to a persons desire to engage in physically risky activities such as ris ky sports and the enjoyment of frightening experiences (Zuckerm an, 1986). Disinhibition represents a form of sensation seeking
63 more related to impulsivity or a lack of self-control. Unfortunately to date, disinhibition, impulsivity, and self-control are often evaluated and discussed sepa rately. The minor differences between these concepts have aris en due to differences in theore tical orientation. For instance, the hierarchical model of personality defines impu lsivity as a biologically based personality trait characterized by acting on impulse, nonplanning, liveliness, and risk -taking (Acton, 2003; Eysenck & Eysenck, 1985). As discussed previo usly, sensation seeking also represents a biologically based personality tra it characterized by a tendency to seek exciting experiences with the disinhibition subscale representing a tendency to act without regard fo r consequences (Acton, 2003). Self-control is considered a behavioral style which is lear ned early in life and is highly resistant to change. High self-c ontrol is characterized by an ab ility to make decisions about current behavior based on considering the long-ter m consequences of that behavior. Individuals with low self-control act without consideration of future conse quences (Gottfredson & Hirschi, 1990). While there are some relatively minor theoretical distinctions in the origins and definitions of disinhibition, impulsivity, and se lf-control, they are all clearl y related by a tendency to act on impulse, without planning for or taking into acco unt the future consequences of ones actions. There is little debate among re searchers regarding the concep tual similarity among these constructs. In part, the distin ction between these cons tructs may be resolv ed via developmental approaches to temperament and personality. An individual may be bor n with a biological disposition towards impulsivity; however, durin g early childhood, parenting behaviors will either encourage or discourage the development of self-control. Background research. In studies evaluating the asso ciation between sensation seeking and substance use, thrill and adventure seeking and disinhibition ar e continually identified as the
64 subscales most strongly related to drug use (A ndrucci, Archer, Pancoast, & Gordon, 1989; Bardo et al., 1996; Donohew, Lorch, & Palmgreen, 1991; Donohew et al., 1999; Kopstein, Crum, Celentano, & Martin, 2001; Zuckerman, et al., 1990). Individuals with higher levels of thrill and adventure seeking and disinhibition have been iden tified by numerous studies to be more likely to initiate substance use, as we ll as have higher levels of use for both specific substances and combined measures of substan ce use (Ball, 1995; Newcomb & Mc Gee, 1989; Zuckerman, 2007). A separate literature has elaborated on a sim ilar connection between the thrill and adventure seeking and disinhibition sensat ion seeking subscales and the de velopment of aggressive and delinquent behaviors (Newcomb & McGee, 1991). More recent studi es have evaluated drug use, aggression, and delinquency combined represen ting general deviance in multiple domains. These studies also find high levels of thrill and adventure seeking a nd disinhibition to be associated with general deviance (Newcomb & McGee, 1991). Other studies have found positive correlations between sensation seeking a nd drug use, aggression and delinquency but are limited in that they utilized cross sectiona l samples of college students (Huba, Newcomb, & Bentler, 1981; White, Labouvie, & Bates 1985). He nce, generalizability to the development of these behaviors over time in middle school is limited. There is some evidence that the effects of th e thrill/adventure seeking and the disinhibition constructs may differ by gender depending on outcome (Newcomb & McGee, 1991). In addition, males are more likely than females to be high sensation seekers. There is also some evidence to suggest that the asso ciation of sensation seeking and drug use may differ when race (comparing Black and White adolescents) is ev aluated. Brown, Miller, & Clayton (2004) found sensation seeking to be predictive of dr ug use among White adolescents but for Black adolescents findings were either non-significant or in the opposite direction than hypothesized.
65 Aside from the findings of Brow n, Miller, & Clayton (2004), evid ence indicates that sensation seeking, in particular the thrill /adventure seeking and disinhibition subscales, are associated with substance use, aggression, and delinquency when evaluated as independent constructs or combined to form a general deviance construct. Ho wever, to date, the role of these subscales in a longitudinal analysis of gender specific a nd common pathways between substance use and aggression and delinquency during th e middle school years has yet to be determined. It is also clear that race/ethnicity shoul d be evaluated to clarify it s role in these pathways. Peer Factors The formation of healthy peer relationships is a key element of current drug use and violence prevention programming (e.g., Dishi on & Loeber, 1985; Kandel & Andrews, 1987). Many of the more effective programs incorporat e tasks aimed at developing healthy peer relationships and effectively a voiding developing friendships with peers who use drugs or engage in delinquent behaviors. Given higher rates of delin quency and aggression overall compared to overall drug use during the middle sc hool years, it follows that more middle school students will develop delinquent peer relationships, than wi ll develop drug using peer relationships. Measurement considerations. Evaluating peer influences on problem behaviors is subject to many challenges. In itial research on the influence of peers focused on the gender composition of ones peer group, with male p eer groups being more conducive to problem behaviors than groups largely composed of fema les (Moffitt et al., 2001). Another way peer groups have been conceptualized is by age of p eers. During adolescence, associating with older peers increases the risk of an individual engagi ng in problem behaviors (Moffitt et al., 2001). Despite the fact that in both of these situati ons peer deviancy was not directly assessed, it is assumed that peer groups composed of mostly males engage in or at least encourage deviant
66 behavior to a greater degree than peer groups largely composed of females. Similarly, since overall rates of deviant behavior increase with age, older peer groups will be more likely to engage in or encourage deviant behavior compared to peer groups of a comparable age to the adolescent. There are other definitional issues to consider when ev aluating the deviancy of an individuals peer group. Soci al learning theory (Bandura & Walters, 1959) postulates modeling of peer behavior as a mechanism that drives the increase in problem behaviors. Based on this premise, researchers generally ope rationally define peer deviancy by assessing peer drug use or by assessing peer engagement in delinquent beha viors depending upon the outcome of interest in their study. However, given th e co-occurrence of substance us e, aggression, and delinquency during adolescence it is important to evaluate both the substance use and delinquent behavior of an individuals peers within the same study. This will provide a more complete picture of the role of peer influences on a dolescent problem behavior. Beyond definitional issues, measurement issues also arise in the evaluation of peer deviancy. As was discussed previously, the most common method of assessment used in the evaluation of problem behaviors is adolescent self report. Alo ng with the problems associated with self reporting ones own deviant behaviors, additional considerations must be taken into account when adolescents report on th e behavior of their peers. Fo r instance, adolescents tend to report that their friends behavior is more similar to their own behavior than it actually is in reality (Wilcox & Udry, 1986). One method recommended by Aseltine (1995) to address this bias involves obtaining information about peer behaviors from actua l peer reports. This can be achieved by obtaining adolescent se lf report on their own devian t behaviors and then asking them to list the names of their closest friends. With this information it is possible to match
67 adolescent surveys with the surveys completed by th eir friends. Peer deviant behavior is then determined based on the adolescents friends repor ts of their own deviant behavior rather than the adolescents perception of their friends behavior. Ho wever, this does require high participation rates within schools as well as fr iend networks within schools. The National Longitudinal Study of Adolescent Health (Add Health) dataset is one example of a large longitudinal dataset with the capacity for this type of measurement technique. Background research. There is substantial evidence of a strong influence of deviant peer association on adolescent antiso cial behavior (Dishion & Patt erson, 2006; Elliot, Huizinga, & Ageton, 1985; Haviland, Nagin, & Rosenbaum, 2007). In a longitudinal anal ysis of a nationally representative sample, Elliot and colleagues (1985) found that deviant peer association predicted subsequent delinquent behavior, even after c ontrolling for previous delinquency. This finding held for both males and females. Similarly, research on the impact of gang affiliation consistently shows increases in delinquent and cr iminal behavior upon joining a gang even after accounting for an individuals hi story of delinquent behavior (H aviland et al., 2007). The impact of deviant peer affiliation on adolescent substance use is highlighted by numerous studies revealing deviant peer affiliation to be one of the strongest predic tors of adolescent substance use as well as relapse after substance abuse treatment (Brown, 1993; Dishion & Patterson, 2006; Friedman & Glassman, 2000; Mayes & Suchman, 2006). While previous research has found a positive association between delinquent friends and adolescent delinquent behaviors (Elliot, Huizinga, & Ageton, 1985; Lynne, Graber, Nichols, Brooks-Gunn, & Botvin, 2007; Moffitt et al., 2001), less work has been done examining the role of friend delinquency on the development of subs tance use, or on the association between friend drug use and the development of aggression, delinque ncy, and drug use. In a prior collaborative
68 study, rates of friend delinquency were much hi gher than friend drug use among urban minority adolescents (Graber, Lynne, Nichols, Brooks-G unn, & Botvin, in preparation). Yet, the differential impact on the development of aggres sion, delinquency and drug use of associating with delinquent peers, as opposed to drug using peers, has yet to be thoroughly evaluated. Hence, the examination of interconnections between the development of aggressive and delinquent behaviors with drug using behaviors while evaluati ng friend delinquency and friend drug use separately as predictors of these pathwa ys will further inform prevention programming. Consistent with research indicating that males tend to engage in higher rates of delinquency and aggression than females, previo us research has found that, among middle school students, male peer groups are more conducive to aggressive and delinquent behaviors (Moffitt et al., 2001). In addition, ther e is some evidence that adolescent males are influenced by association with delinquent or drug using friends more so than females (Erickson, Crosnoe, & Dornbusch, 2000). However, differential associa tion or susceptibility to delinquent peers does not completely explain the gender difference in a ggressive and delinquent be haviors. As such, it is of value to examine gender differences in pathways to drug use that incorporate interconnections with aggression and delinquency. Unfortunately, there is very li ttle empirical information regarding the influence of deviant peers among ethnically diverse samples. Only one study that we know of has evaluated differences in the influence of deviant peers among different ethniciti es. This study provided some evidence that friend's substance use opera tes differently when comparing White and Black adolescents. Black adolescents with friends who smoked cigarettes in the 6th grade were less likely to smoke themselves in 10th grade suggesting that friend's us e is not directly related to Black youths' own use of drugs (Brown, Miller, & Clayton, 2004). Clearly, further research is
69 needed evaluating the influence of devian t peers on multiple antisocial behaviors among ethnically diverse samples. Referring to Bronfenbrenners ( 1979) ecological model, it is im portant to evaluate personcontext interactions regarding th e pathways to problem behaviors. Initiation of substance use often occurs with friends who ar e using drugs; however, it is clear that exposure to drug using friends does not operate independent of the i ndividual who chose those friends (Newcomb & Bentler, 1989). Adolescents often select frie nds with common interests based on their own psychological characteristics. Yet, no studies have examined interact ions between sensation seeking and deviant peer associa tion. The proposed study plans to examine this interaction and evaluate it in regards to pa thways to problem behaviors.
70 CHAPTER 3 SPECIFIC AIMS While numerous studies have evaluated initiati on and use of drugs and alcohol as well as increases in rates of aggressi on and delinquency during the middle school years, interconnections between these factors have not systematically been evaluated longitudinally in a large sample of urban, minority adolescents. St udies have shown that rates of aggression and delinquency begin to increase earlier than rates of drug use, and co-occurrence of th ese behaviors is often reported. However, the temporal associations among thes e factors and individual differences in these pathways have yet to be fully evaluated. It is also important to examine the role of indi vidual and contextual factors as predictors of these negative adjustment outcomes. Sensation seeking, specifically thri ll and adventure seeking and disinhibition (i.e., impulsivity, lack of self-control), represen ts an individual characteristic that has been identified as sa lient to engagement in drug use and some delinquent behaviors. Similarly, contextual factors such as friend delinquency and friend drug us e have been shown to be related to aggression, delinque ncy, and drug use when evaluate d separately for each outcome but have yet to be studied as pr edictors of interconnections be tween these outcomes. Moreover, pathways between these factors a nd substance use may be mediated by engagement in aggressive or delinquent behaviors. In addition, there is substantia l evidence indicating that rates of engagement in aggression, delinquency, and drug use differ by gender and race/ ethnicity However, more recent research seems to indicate that the gender gap is narrowing, especially in regards to aggression and some types of substance use. It is therefore imperative to examine commonalities as well as differences between genders and races/ethnicities regard ing pathways to aggression, delinquency, and drug use in middle school. Like wise, sensation seeking and the impact of
71 deviant peers may vary by gender. Hence, these constructs may be particularly informative to understanding factors associated with gender specif ic pathways to drug use. Prevention efforts (and interventions) largely focus on reducing the rates of all thr ee of these negative outcomes among adolescents. These efforts would greatly benefit from a clearer understanding of both gender specific and common pathways to drug use, incorporating interconnections with pathways to aggression and delinquency. Based on previous research, it was hypothesized that significant increases in antisocial behavior s would be observed as well as gender and racial/ethnic differences in average levels of anti social behaviors. However, the primary aims of the current study were not necessarily hypothesis driven but rather, were more exploratory, stemming from previous theoretical work outlined above. The specific aims of this study were as follows: 1. The first aim examined differences in patter ns of change for aggression, delinquency, and substance use for girls, boys, or both genders Bi-directional or te mporal associations between these patterns of change in substa nce use, aggression, and delinquency were evaluated over three assessments in early adolescence. 2. The second aim evaluated the mediating role of individual changes in aggression and delinquency on associations between substance use, sensation seeking, and peer deviancy for girls, boys, or both genders. Two components of general sensation seeking, thrill/adventure seeking and disinhibition, were evaluate d along with the general sensation seeking construct. Two aspects of peer deviancy were also evaluated (peer delinquency and peer drug use). 3. The third aim evaluated the moderating role of gender on associati ons between antisocial behaviors (aggression, delinquency, and substa nce use), sensation seeking, and peer deviancy. Two components of general sens ation seeking, thrill/adventure seeking and disinhibition, were evaluated along with the general sensation seeking construct. Two aspects of peer deviancy were also evalua ted (peer delinquency and peer drug use) for gender differences in associations with a ggression, delinquency, and substance use. 4. The fourth aim evaluated the moderating role of race/ethnicity on associations between antisocial behaviors (aggressi on, delinquency, and substance us e), sensation seeking, and peer deviancy. Two components of general sensation seeking, thrill/adventure seeking and disinhibition, were evaluate d along with the general sens ation seeking construct. Two aspects of peer deviancy were also evaluated (peer delinquency and peer drug use)
72 for race/ethnic differences in associations with aggression, delinquency, and substance use. These research questions were addressed via se condary data analysis of data collected as part of a large, school-based violence prevention program evaluation, based on the LST Program for alcohol, tobacco, and drug preven tion. Other data sets might be used to examine these issues. In particular the ADD Health public domain data set includes many of the constructs of interest in the present project; in addition, ADD Health is a nationally representative sample. However, the violence prevention data set to be used in this project focuses intensively on the middle school years with an urban minority sample. The dearth of information on pathways to drug use and delinquency among this group of adolescents validates the choice of th is particular dataset and results of the proposed study will contribute greatly to the field.
73 CHAPTER 4 RESEARCH DESIGN AND METHODS Design A total of 42 public and parochial middle schoo ls in New York City participated in the evaluation study. All schools partic ipated in baseline data coll ection activities with their 6th grade classes, prior to the intervention, and annua l surveys in 7th and 8th grades; half the schools received prevention programming for three years. Only participants assigned to the control condition at baseline wi ll be used in the proposed study in order to avoid contamination with potential intervention effects. Participants Participants were 2,931 young adolescents draw n from the control condition of the LST program evaluation study. In the 6th grade, participants of the study reported a mean age of 11.72 years (SD = 0.54) with a range from 9.64 to 14 years. Fifty percent of the sample is female and the sample is largely minority wi th 48% Black, 30% Latino, 7% White, 5% Asian, and 9% Other (1% didnt report race). Just under ha lf of the students came from an intact family (43%), 32% lived with a single parent, 14% lived in blended families (with stepparents or split time between mother and fathers home) and 6% lived in househol ds without any parent present (with other relatives, or with foster parents or guardians). Although a measure of family SES was not available, archival public school records of participating schools showed that the majority (88%) of schools had greater than 65% student eligibility for free or reduced lunch. Youth were enrolled in public ( 90%) and parochial (10%) schools.
74 Procedure A passive consent procedure approved by Weill Cornell Medical Colleges IRB was used to inform parents about the nature of the st udy and to provide them with an opportunity to disallow their childs participation. A consent form describing the st udy and the self-report survey was distributed in the schools, as well as mailed directly to students homes. Students whose parents indicated they did not want them to participate in the self-report survey did not complete any of the data collection activities. Data collectors for survey data collection we re recruited, trained, and supervised by the Project Manager at Teachers College, Columbia University under the supervision of the PI and Co-PIs. Equal numbers of males and females, as well as Black and Latino college students were recruited from City and Community college s and trained for the school based study. A standardized and effective system for traini ng data collectors was developed under the NIDA center grant. Data collectors were organized into teams of 5 and were supervised by a field leader. Each team was able to complete the data collection for student s in a single day per school. The primary collection of survey data to ok about 2 weeks at each school. Smaller teams of two data collectors returned to schools on at least three occasions after each scheduled data collection to collect data on absentees. Quality control in the collection of the data was assured by careful training of all data collection staff, use of carefully written protocols, and by regula r observation of field procedures by the investigators. Data collectors observed a training video of data collection protocols in addition to signing a Confidentiality Pledge regarding the handling of sensitive participant information. The survey was divided into two booklets a nd data collection was conducted on two days during regular 40-minute class periods. A multi-ethnic team of three to five data collectors
75 administered the questionnaire following a standardiz ed protocol used in pr evious research (e.g., Botvin, Schinke, Epstein, & Diaz, 1994). To ensure the quality of self-report data, identification codes rather than names were used to emphasize th e confidential nature of the questionnaire and students were assured about the c onfidentiality of thei r responses. Carbon monoxide (CO) breath samples were also collected at 6th, 7th, and 8th grade to enhance the validity of self-report data utilizing a variant of the bogus pipeline pro cedure developed by Evans and his colleagues (Evans, Hansen, & Mittlemark, 1977). While this measure was used to in crease the validity of questions pertaining to cigare tte smoking, studies have shown bogus pipeline procedures can also increase the validity of reporting on other problem behaviors (Tourangeau, Smith, & Rasinski, 1997). Measures Surveys were used to assess aggressive and delinquent behaviors as well as drug, alcohol, and tobacco use in each year of data collection. In addition, su rveys assessed numerous factors that have been identified in prio r studies as correlates of these be haviors. The measures used in this investigation have been developed from in depth pilot testing and us e of measures in large school-based drug prevention studies with ur ban multi-ethnic youth (See Appendix A). All measures were collected a nnually in the spring of 6th, 7th, and 8th grades. Demographics Data concerning the demographic characteristic s of the participants were collected using standard survey items concerning gender, age, family structure, race/ethnicity, socioeconomic status (i.e., receive free or reduced-price schoo l lunch). For purposes of analyses, a single dichotomous variable was created to capture th e type of household stru cture where 1 indicates living in a two-parent household. Students from single and no-parent households were coded as 0. Two dichotomous variables were created to capt ure participants race or ethnic affiliation.
76 The first represented participants who were La tino (coded one) versus all other participants (coded zero). The second represented participants who were Black/African American (coded one) versus all other participants (coded zero) Values of zero on both of these variables represented individuals from a ll other ethnic groups (White/Caucasian, Asian, American Indian, and Other). A single dichotomous variable wa s created to indicate if the adolescent was attending a public or a parochial school (where 1 represents particip ants who attended public school and 0 for parochial school students). Drug Use Survey items assessed frequency of cigare tte smoking, drinking alcohol, drinking until drunk, smoking marijuana, smoking marijuana un til high or stoned, and using inhalants. Frequency of each item was measured on a nine-point scale with the following choices: never, a few times but not in the past year, a few times a year, once a month, a few times a month, once a week, once a day, or more than once a day. Due to the fact that reports of individual substance use is quite low at the initial assessment but increases, the pr oposed study plans to create a composite sum score of drug use, such that higher values represent more overall drug use at each grade. Cronbachs alpha is somewhat low in the 6th grade ( = .61) due to the low reported rates of substance use at this age, but incr eases to appropriate levels in the 7th ( = .78) and 8th ( = .86) grades. Previous studies have found compos ites to be useful, although there are limitations such as an inability to test gateway models (Chen & Kandel, 1995; Kandel & Logan, 1984). The present study will not test this model. Aggression Aggression was assessed via the aggression scale of the Youth Self-Report (YSR, Achenback & Edelbrock, 1986). Students were asked how many times in the past month they had engaged in ten incidents of overt ly aggressive behavior in the 6th grade ( = .93), 7th grade (
77 = .94), and 8th grade ( = .94). Items included Yelled at someone (you were mad at), Told someone off, Pushed or shoved someone on purpose, and Hit someone. Response categories were on a 5-point scale. Response options included 1 ( Never ), 2 ( Once), 3 ( 2-3 times ), 4 ( 4-5 times), and 5 ( More than 5 times). Items were rescored onto a scale of 0-4 and then summed to create a continuous measure where highe r scores indicate greater aggression. The response options were changed from the 3-point s cale in the YSR to the 5-point scale in this study to be consistent with the other measures in the survey. Delinquency Students were also asked how many times in the past year they had engaged in ten incidents of delinquent behavior (adapted from Elliott, Huizin ga, & Menard, 1989). The same ten items were used in the 6th grade ( = .86), 7th grade ( = .88), and 8th grade ( = .90). Two separate subscales were created to dis tinguish violent delinquency (6 items; 6th grade = .81, 7th grade = .83, and 8th grade = .85) and non-violent delinquency (4 items; 6th grade = .69, 7th grade = .77, and 8th grade = .80). Items for violent delinquency included Thrown objects such as rocks or bottles at cars or people, Picked a fight with someone, Hit someone with the idea of seriously hurting them, Taken someth ing from a person by force (other than just playing around), Beat up on someone or f ought someone physically if they provoked you (other than just playing around), and Taken part in a fight where a group of your friends were against another group. Items for non-violent delinquency included Purposefully damaged or destroyed property or things that did not belong to you, Taken something from a store when the clerk wasnt looking, Int entionally damaged or messed up something in a school or some other building, and Taken something worth less than $50 that did not belong to you. Response categories were on a 5-point sc ale. Response options included 1 ( Never), 2 ( Once), 3 ( 2-3 times), 4 ( 4-5 times), and 5 ( More than 5 times). Within each subscale, items were rescored
78 onto a scale of 0-4 and then summed to create a continuous measure where higher scores indicate greater violent or non-violent delinquency. Friend Delinquency Students were asked to indicat e how many of their friends had engaged in seven incidents of delinquent behavior in the past year. Th e same seven items were used to assess friend delinquency in the 6th grade ( = .88), 7th grade ( = .91), and 8th grade ( = .92). These were taken from a violence/delinquency scale devel oped by Elliot et al. (1989). Items for friend delinquency included Hit or threat ened to hit someone without any real reason, Beat someone or fought someone physically if they were provok ed (other than just playing around), Ruined or damaged something on purpose that wasnt th eirs, and Stolen something worth less than $50. Response options included 1 ( None ), 2 ( Less than half), 3 ( About half ), 4 ( More than half ), and 5 ( All or almost all ). Items were rescored onto a scale of 0-4 and then summed to create a continuous measure where higher sc ores indicate associating with more friends who engage in a higher number of delinquent behaviors for each time point. Friend Drug Use Adolescents reported how many of their friends had tried tobacco, al cohol, marijuana, inhalants, or other drugs. The same five it ems were used to assess friend drug use in the 6th grade ( = .79), 7th grade ( = .80), and 8th grade ( = .83). Response options included 1 ( None ), 2 ( Less than half ), 3 ( About half ), 4 ( More than half), and 5 ( All or almost all ). Items were rescored onto a scale of 0-4 and then summed to create a continuous measure where higher scores indicate associating with a greater number of substance using friends for each time point. Perceived reports of friend delinquency and drug use have been used extensively in the past. Previous studies have also attempted to validate a dolescent perceptions of peer delinquency and drug use by assessing rates of th ese behaviors directly from the adolescents
79 friends, rather than having the adolescent report on their perception of their friends behaviors only. However, it is not possible in the present investigation to match adolescent reports to peer reports of delinquent and dr ug using behavior due to th e nature of the dataset. Sensation Seeking Thrill and adventure seeking. Students were asked to indicate how much they agree with four statements regarding enjoym ent of risky activities in the 6th grade ( = .79), 7th grade ( = .82), and 8th grade ( = .82). Items for thrill and advent ure seeking included I enjoy taking risks, I would enjoy fast driv ing, I would do almost anythi ng on a dare, and I think life with no danger in it would be dull for me. Responses were measured on five-point scales ranging from strongly disagree to strongly agree. The validity of this four item measure of thrill and adventure s eeking was confirmed in a study conducted by Eysenck & Eysenck (1975). Items were rescored onto a scale of 0-4 and th en summed to create a continuous measure where higher scores indicate greater en joyment of risky activities. Disinhibition. Students were asked to indicate how much they agree with 13 statements regarding the ability to focus and inhibit certain behaviors in the 6th grade ( = .79), 7th grade ( = .80), and 8th grade ( = .80). This measure was modeled after a scale developed by Rosenbaum (1980). Items for disi nhibition included I am easily di stracted from my work, It doesnt really take much to calm me down when I am excited or all wound up (reverse coded), and I have been told that I interrupt people in c onversations. Responses were measured on five-point scales ranging from s trongly disagree to strongly agree. Items were rescored onto a scale of 0-4 and then summed to create a con tinuous measure where higher scores indicate less ability to focus and inhibit behaviors. A composite measure of sensation seeki ng was formed by combining the thrill and adventure seeking subscale and the disinhibition subscale. This composite measure was formed
80 by summing the two subscales to create a continuous measure where higher scores indicated a greater enjoyment of risky activities and less ab ility to inhibit behavi ors while lower scores indicated a dislike of risky activ ities and a greater ability to fo cus and inhibit behaviors. Tracking and Attrition Participants were maintained over the three y ears of data collection by using a combination of existing tracking information from the investig ators own files or provided to them by middle school administrators and (for New York City schools) information obtained from the New York City Board of Education Office of System s Development Support (OSIS) database (a computerized biofile on every studen t ever enrolled in the system). Several sources of information were used to track students throughout the course of the study including: (1) ID information obtained at the beginning of the study (name, ID code, school, home room, grade); (2) parent information collected annually on tracking cards and (3) school records (e.g., name of parent(s) or guar dian, home address, phone number). Unique student identification numbers we re linked to students name, current address, phone number, current school, and enrollment status. Those numbers were used to track students throughout their academic careers and provided an ex cellent method of locating students. At the first assessment in 6th grade, 2931 students participat ed in the control condition of the larger study. From 6th to 7th grade, 5% of the sample was lost to attrition with an additional 31% lost between 7th and 8th grades resulting in a longitudi nal sample of 1922 students. Analyses of attrition bias were condu cted to test for differences in 6th grade between students in the current study and those who were dropped from any of the planned analyses (t-tests for continuous variables, 2 tests on background variables).
81 Analysis Plan Data analysis encompasses two main phases in the proposed study: ps ychometric analysis followed by analysis of specific research questions Initial data cleaning and file construction were conducted at the Institute for Prevention Re search at Weill Medical College of Cornell University which was responsible for all data management Power Analysis For Ordinary Least Squares Regression (OLS) models, Cohen and Cohen (1983) recommend a power level of .7 to .9 for determini ng necessary sample sizes for research. Given conservative estimated effect sizes of .25 to .35 (Cohen & Cohen, 1983), an alpha level of .05, a power level of .9 (90% chance an effect will be found if it is there), and an N of 175, as many as 20 independent variables can be used in hierarch ical regressions and pa th analyses (including control and mediator/moderator variables). In this study, the sample size was 2,931 in the first assessment (50% female). As reported previous ly, a longitudinal sample of approximately 1,922 participants were maintained (53% female). In addition, mean levels of behaviors may differ by race/ethnicity, but such differences would not influence the extent th at underlying processes accounting for behavioral outcomes would be similar. All analyses controlled for ethnic/racial group with the exception of those for Aim 4, which specifically examined the possible moderating effects of race/ethnic gr oup. Results of this power analysis reveal that the sample size is more than adequate to test the research que stions of interest in the proposed study. Nagin (1999) has conducted group-based trajec tory analyses with samples of 411 and 1,037. Thus, it is likely that power was sufficient in this study for these analyses. Across the aims, the initial sample size of 2,931 affords te sting many complex models; however, attention will be paid to violati on of norms when choosing specific esti mation methods and test statistics.
82 Additionally, some of the methods used in this study are particularly useful because of their robustness against missing data. Longitudinal Analyses Because constructs assessed in the surveys were measured at 3 points in time, it was expected that change over time, or trajectories of constructs were more important than specific levels at any given time of assessment. As such, major research objectives outlined in Aim 1 were pursued via innovative methods developed to assess developmental trajectories for groupbased rather than individual growth curves (Nagin, 1999; Nagin & Tremblay, 2001). This method was used to identify di stinctive groups of i ndividual trajectories within samples with multi-wave data. This method includes the capability to: (1) relate group membership probability to individual characteristics and circumstances; (2) use the group membership probabilities to create profiles of group members; (3) add time-varying cova riates to trajectory models; and (4) estimate joint trajectory models of distinct but related be haviors. The advantage of group-based trajectory modeling is the ability to identify distinct groups of trajectories within the population. This is in contra st to hierarchical and latent growth curve modeling, both of which assume a continuous distribution of traject ories within a population. This is in direct opposition to the idea of distinct clusters or ca tegories of development and therefore it is awkward to use these methods when evaluating re search questions that address developmental trajectories that are inherently categorical. Testing Mediating and Moderating Effects Mediating effects outlined in Aim 2 were evaluated through a series of HLM equations where both the predictors and the mediators were person-mean centered to assess the impact of individual changes in aggression, delinquency, sensation seeking, and peer deviancy on substance use. Each of the aforementione d variables was person-mean centered by first
83 estimating each participants personal average on a particular variable ac ross time. For example, an individuals self-reported aggression in the 6th, 7th, and 8th grades was averaged resulting in an average level of aggression sp ecific to that individual duri ng middle school. Next, each participants personal average on a particular variable was subtracted from the time specific assessments for that variable. In other words, an individuals average level of aggression was subtracted from their self -reported aggression in 6th, 7th, and 8th grades. The resulting values provided information about engagement in a speci fic behavior at higher or lower rates than typical for a particul ar individual. Examination of gender as a moderator of associations was tested under Aim 3 while the moderating effect of race/et hnicity was tested under Aim 4 HLM was used for testing crosslevel interactions invo lving gender (level 2 predictor) and sensation seeking or deviant peer association (level 1 predictors) on the dependent variable of interest The level 1 predictors were person-mean centered to assess the effect of ch anges in sensation seek ing or deviant peer association within an individua l on the dependent variable of interest rather than average associations between these variable s across the sample. Analysis of Specific Aims This study addresses 4 primary aims. Aim 1 examined differences in patterns of change for aggression, delinquency, and substa nce use for girls, boys, or bot h genders. Bi-directional or temporal associations between these patterns of change in substance use, aggression, and delinquency were evaluated over three assessments in early adolescence. This aim was addressed by initial examinations of group-ba sed trajectories for each outcome individually followed by dual trajectory analyses. The initia l group-based trajectory an alysis determined if there were different patterns of change within each construct. The dual trajectory analysis evaluated patterns of cha nge within two outcomes of interest simultaneously as well as estimated
84 the probability that an individual following a particular trajectory for outcome A (e.g., aggression) also followed a pa rticular trajectory on outcome B (e.g., delinquency). Gender differences in trajectory group me mbership were analyzed. Aim 2 evaluated the mediating role of indivi dual changes in aggre ssion and delinquency on associations between substance use, sensation seek ing, and peer deviancy for girls, boys, or both genders. Two components of general sens ation seeking, thrill/adventure seeking and disinhibition, were evaluated along with the general sensation seeki ng construct. Two aspects of peer deviancy were also evaluated (peer delinque ncy and peer drug use). The first step in the mediation analysis was to determine if there wa s a significant associati on between the predictor (e.g., sensation seeking) and substa nce use. Next, the associati on between the predictor and the mediator (e.g., aggression or delinquency) was evaluated for significance. Finally, both the predictor and the mediator were simultaneousl y evaluated for significan t associations with substance use. Aggression or delinquency served as a mediator if the previously significant association between the predicto r and substance use was either rendered non-significant or was significantly reduced. These series of analyses were conducted using HLM with both the predictor variables and the mediat or variables person-mean centere d to assess changes within an individual. An analysis of the interaction between sens ation seeking and peer deviancy was first evaluated as the predicto r. In the event of a non-significan t interaction, analyses proceeded by evaluating the main effects of general sensation seeking, the two sub-dimensions of sensation seeking, and the peer deviancy va riables as the predictors. Aim 3 evaluated the moderating role of ge nder on associations between antisocial behaviors (aggression, delinquency, and substance use), sensation seeking, and peer deviancy. Two components of general sens ation seeking, th rill/adventure seeking and disinhibition, were
85 evaluated along with the general sensation seeki ng construct. Two aspects of peer deviancy were also evaluated (peer delinque ncy and peer drug use) for gende r differences in associations with aggression, delinquency, and substance use. As discussed previously, HLM was used to test cross-level interactions i nvolving gender (level 2 predictor) and sensation seeking or deviant peer association (level 1 predic tors) on aggression, delinquency, and substance use. The level 1 predictors were person-mean cente red to assess the effect of cha nges in the variables within an individual on the outcomes of interest. First, the moderating role of ge nder on the interaction between sensation seeking and peer deviancy was evaluated. In the ev ent of a non-significant three-way interaction, gender was evaluated as a moderator of the a ssociation between the outcomes of interest and the main effects of ge neral sensation seeking, the two sub-dimensions of sensation seeking, and the pe er deviancy variables. Aim 4 evaluated the moderating role of race/ethn icity on associations between antisocial behaviors (aggression, delinquency, and substance use), sensation seeking, and peer deviancy. Both components of general sensation seeking, thrill/adventure seeki ng and disinhibition, were evaluated along with the general sensation seeki ng construct and two aspects of peer deviancy (peer delinquency and peer drug use) for race/ethni c differences in associat ions with aggression, delinquency, and substance use. Similar to the moderation analyses of the previous aim, HLM was used to test cross-level inte ractions involving race/ethnicity (level 2 predictor) and sensation seeking or deviant peer asso ciation (level 1 predictors ) on aggression, delinquency, and substance use. The level 1 predictors were pers on-mean centered to assess the effect of changes in the variables within an indivi dual on the outcomes of interest. First, the moderating role of race/ethnicity on the interaction between sensatio n seeking and peer deviancy was evaluated. If this three way interaction was not significant, ra ce/ethnicity was evaluated as a moderator of the
86 association between the outcomes of interest and the main effects of general sensation seeking, the two sub-dimensions of sensation seek ing, and the peer deviancy variables.
87 CHAPTER 5 RESULTS Descriptive Analysis Gender and racial/ethnic differe nces in demographic, outcome and predictor variables as well as overall sample statistics for variables used in this study are presented in Table 5-1. There were no significant differences in the percen tage of males compared to females among the demographic variables (e.g., race/ethnicity, house hold structure, and school type). However, there were significant racial/ethnic differences regarding household structure and school type. Black adolescents were less like to live with two parents in the 6th grade compared to Latinos and White/Other adolescents, 2(2, N = 2,879) = 112.50, p < .001. Black and Latino adolescents were also more likely to attend public school as opposed to private school compared with White/Other adolescents, 2(2, N = 2,912) = 22.12, p < .001. Table 5-1. Descriptive statistics Gender Race / Ethnicity Males Females Black Latino White/Other Total sample Demographic Variables Gender (% female) 53% 50% 47% 50% % African American 46% 50% 48% % Latino 31% 30% 30% % White/Other 23% 20% 22% % Living with Two Parents 57% 56% 47% 62% 71% 57% School Type (% public) 90% 90% 92% 90% 86% 90%
88 Table 5-1. Continued. Gender Race / Ethnicity Males Females Black Latino White/Other Total sample Outcome Variables M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) Aggression 6th 13.48 (11.02) 11.62 (9.95) 14.06 (10.78) 10.49 (9.87) 12.05 (10.40) 12.54 (10.54) Aggression 7th 18.76 (12.44) 19.17 (11.75) 20.35 (12.22) 17.62 (11.75) 17.34 (11.74) 18.98 (12.08) Aggression 8th 19.10 (12.80) 19.64 (12.09) 21.05 (12.54) 17.22 (12.07) 18.80 (12.16) 19.39 (12.42) Delinquency 6th 4.82 (6.40) 2.84 (4.32) 4.39 (5.89) 3.21 (4.94) 3.43 (5.44) 3.83 (5.55) Delinquency 7th 6.99 (8.10) 5.43 (6.58) 6.76 (7.77) 5.82 (6.93) 5.11 (6.75) 6.16 (7.37) Delinquency 8th 7.54 (8.87) 5.86 (7.42) 7.38 (8.75) 5.71 (7.27) 6.34 (7.94) 6.64 (8.17) Substance Use 6th 0.62 (1.78) 0.38 (1.21) 0.49 (1.47) 0.47 (1.39) 0.58 (1.81) 0.50 (1.52) Substance Use 7th 0.90 (2.00) 0.95 (2.18) 0.86 (1.99) 1.06 (2.24) 0.91 (2.19) 0.93 (2.10) Substance Use 8th 1.21 (2.63) 1.37 (2.83) 1.25 (2.58) 1.33 (2.66) 1.33 (3.14) 1.30 (2.74) Predictor Variables M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) Friend Delinquency 6th 4.49 (5.63) 3.88 (4.99) 4.58 (5.56) 3.77 (5.00) 3.87 (5.19) 4.18 (5.33) Friend Delinquency 7th 6.36 (6.78) 5.97 (6.36) 6.59 (6.90) 5.89 (6.11) 5.39 (6.19) 6.15 (6.56) Friend Delinquency 8th 6.59 (6.79) 6.02 (6.63) 6.78 (7.04) 5.73 (6.22) 5.99 (6.57) 6.29 (6.71) Friend Drug Use 6th 1.36 (2.66) 1.26 (2.47) 1.39 (2.70) 1.37 (2.55) 1.06 (2.31) 1.31 (2.58) Friend Drug Use 7th 2.20 (3.13) 2.46 (3.27) 2.29 (3.27) 2.59 (3.27) 2.12 (2.96) 2.34 (3.21) Friend Drug Use 8th 2.76 (3.58) 3.32 (3.87) 3.21 (3.86) 3.12 (3.74) 2.64 (3.48) 3.06 (3.75) Sensation Seeking 6th 3.41 (1.44) 3.13 (1.31) 3.27 (1.39) 3.27 (1.32) 3.26 (1.42) 3.27 (1.38) Sensation Seeking 7th 3.64 (1.37) 3.48 (1.31) 3.47 (1.34) 3.72 (1.29) 3.54 (1.39) 3.55 (1.34) Sensation Seeking 8th 3.70 (1.32) 3.46 (1.27) 3.53 (1.29) 3.63 (1.30) 3.58 (1.32) 3.57 (1.30)
89 Table 5-1. Continued. Gender Race / Ethnicity Males Females Black Latino White/Other Total sample Predictor Variables M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) Adventure Seeking 6th 2.86 (1.15) 2.57 (1.02) 2.72 (1.11) 2.67 (1.07) 2.73 (1.08) 2.71 (1.09) Adventure Seeking 7th 2.98 (1.12) 2.79 (1.04) 2.80 (1.11) 3.00 (1.04) 2.91 (1.05) 2.88 (1.08) Adventure Seeking 8th 3.06 (1.07) 2.80 (1.01) 2.87 (1.04) 2.98 (1.04) 2.96 (1.07) 2.93 (1.05) Disinhibition 6th 3.44 (0.59) 3.43 (0.57) 3.44 (0.57) 3.40 (0.56) 3.46 (0.61) 3.43 (0.58) Disinhibition 7th 3.35 (0.56) 3.31 (0.57) 3.33 (0.55) 3.30 (0.53) 3.38 (0.63) 3.33 (0.56) Disinhibition 8th 3.36 (0.55) 3.35 (0.55) 3.34 (0.56) 3.35 (0.53) 3.39 (0.56) 3.35 (0.55) Significant increases in aggression F (2, 3254) = 396.82, p < .001, 2 = .196, delinquency F (2, 3266) = 178.98, p < .001, 2 = .099, and substance use F (2, 3168) = 115.35, p < .001, 2 = .068, were observed across middle school for the samp le overall. There were a few differences in average rates of these outcomes between gender s and race/ethnicities. There was a significant interaction between grade and gender for aggression, F (2, 3254) = 5.36, p = .005, 2 = .003; however, bonferroni corrected follow-up tests were not significant. The difference between males and females on sixth grade ag gression approached significance (p = .061) with males reporting slightly higher rate s. Males did report significan tly higher rates of delinquency compared to females across middle school, F (1, 1633) = 28.59, p < .001, 2 = .017. There were no gender differences in reports of substance use. Black adolescents reported significantly higher levels of aggression F (2, 1626) = 17.95, p < .001, 2 = .022 and delinquency F (2, 1632) = 9.81, p < .001, 2 = .012 compared to Latino and White/O ther adolescents. There were no significant racial/ethnic differences in reported rates of substance use.
90 Significant increases across middle school were observed in both of the deviant peer predictor variables, delinquent friends F (2, 3218) = 106.78, p < .001, 2 = .062 and drug using friends F (2, 3230) = 214.40, p < .001, 2 = .117. Gender and racial/ethnic difference were also observed. Males reported higher rates of delinquent friends than females, F (1, 1609) = 3.95, p = .047, 2 = .002. Additionally, there was a signifi cant interaction between grade and gender regarding friend drug use, F (2, 3230) = 5.11, p = .006, 2 = .003. Bonferroni corrected followup tests revealed that in 8th grade, females reported signifi cantly more drug using friends compared to males. White/Other adolescents reported significantly fewer delinquent friends F (2, 1608) = 4.98, p = .007, 2 = .006 compared to Black adolescent s as well as significantly fewer drug using friends F (2, 1614) = 5.43, p = .004, 2 = .007 compared to both Black and Latino adolescents. Significant increases across middle school we re also observed for general sensation seeking F (2, 3218) = 106.78, p < .001, 2 = .062. Both subscales showed significant changes as well; thrill/adventure seeking increased over time F (2, 1766) = 37.42, p < .001, 2 = .041 and disinhibition decreased over time F (2, 1984) = 24.42, p < .001, 2 = .024. Males reported significantly higher levels on the meas ure of general sensation seeking F (1, 866) = 8.23, p = .004, 2 = .009 as well as the thri ll/adventure seeking subscale F (1, 883) = 16.45, p < .001, 2 = .018. There were no gender differen ces in levels of disinhibition. Latinos reported significantly higher levels of general sensation seeki ng compared to White/Other adolescents F (1, 865) = 3.33, p = .036, 2 = .008; and there was a sign ificant interacti on between grade and race/ethnicity regarding thrill/adventure seeking F (4, 1764) = 2.72, p = .028, 2 = .006. Bonferroni corrected follow-up tests revealed that La tinos reported significantly highe r levels of thrill/adventure seeking compared to both Black and White/Other adolescents in the 7th grade and compared to
91 Black adolescents only in the 8th grade. There were no race/ethnic differences in 6th grade thrill/adventure seeking or disinhibition at any grade. An evaluation of the correlations among the predictor variables re vealed very strong associations between the overall sensation seeking vari able and the 2 dimensions of sensation seeking (thrill/adventure seeking and disinhibition). There were also very strong correlations between friend delinquency and friend drug use (See Table 5-2). Intermediate analyses determined that the two dimensions of sensa tion seeking did not provi de any new information beyond what was obtained through evaluations of overall sensation seeking. Nor did friend drug use provide any additional information above what was obtained through evaluations of friend delinquency. Given the high correlations betw een these variables and the redundancy in their associations with the outcomes of interest, fu rther analyses are reporte d for overall sensation seeking and friend delinquency only. Table 5-2. Correlations among the predictor variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 Sensation Seeking 6th 2 Sensation Seeking 7th 0.49 3 Sensation Seeking 8th 0.43 0.54 4 Thrill/Adventure Seeking 6th 0.91 0.45 0.40 5 Thrill/Adventure Seeking 7th 0.44 0.91 0.49 0.46 6 Thrill/Adventure Seeking 8th 0.38 0.48 0.91 0.40 0.50 7 Disinhibition 6th -0.65 -0.33 -0.25 -0.29 -0.19 -0.13 8 Disinhibition 7th -0.34 -0.63 -0.36 -0.20 -0.26 -0.19 0.42 9 Disinhibition 8th -0.29 -0.37 -0.63 -0.17 -0.20 -0.25 0.34 0.50
92 Table 5-2. Continued. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 10 Delinquent Friends 6th 0.29 0.20 0.19 0.26 0.18 0.15 -0.19 -0.14 -0.16 11 Delinquent Friends 7th 0.28 0.32 0.24 0.26 0.26 0.18 -0.19 -0.26 -0.23 0.37 12 Delinquent Friends 8th 0.29 0.30 0.37 0.25 0.23 0.32 -0.21 -0.26 -0.27 0.31 0.47 13 Drug Using Friends 6th 0.22 0.17 0.14 0.19 0.14 0.11 -0.17 -0.13 -0.11 0.51 0.21 0.19 14 Drug Using Friends 7th 0.23 0.27 0.20 0.20 0.21 0.13 -0.18 -0.25 -0.22 0.31 0.57 0.35 0.26 15 Drug Using Friends 8th 0.26 0.24 0.28 0.22 0.16 0.23 -0.18 -0.26 -0.22 0.18 0.37 0.61 0.19 0.45 Note : All correlations were significant at the p < .001 level An evaluation of the correlations between th e outcomes of interest revealed high positive correlations between aggression and delinquenc y (See Table 5-3). Both aggression and delinquency showed moderate positive correlations with substance use. Table 5-3. Correlations among outcome variables 1 2 3 4 5 6 7 8 1 Aggression 6th 2 Aggression 7th 0.49 3 Aggression 8th 0.42 0.57 4 Delinquency 6th 0.71 0.35 0.31 5 Delinquency 7th 0.46 0.67 0.44 0.49 6 Delinquency 8th 0.41 0.47 0.66 0.42 0.57 7 Substance Use 6th 0.25 0.11 0.10 0.37 0.20 0.15 8 Substance Use 7th 0.25 0.30 0.18 0.23 0.41 0.27 0.24 9 Substance Use 8th 0.21 0.22 0.26 0.19 0.31 0.37 0.19 0.47 Note : All correlations were significant at the p < .001 level
93 In sum, significant increases in aggression, de linquency, substance use, peer deviancy, and sensation seeking were observed across mi ddle school with some notable gender and racial/ethnic differences in average rates of th ese variables. Males re ported higher levels of delinquency and more delinquent friends than fe males as well as higher levels of general sensation seeking and thrill/adventure seeking. There were no differences in average rates of aggression or substance use between males and females in this sample and females reported more drug using friends than males in 8th grade. Consistent with previous research, Black adolescents reported higher levels of both a ggression and delinquency co mpared to the other racial/ethnic groups in this sample. There were no significant differences between racial/ethnic groups in reported rates of substance use. Black adolescents reporte d significantly more delinquent and drug using friends while Latinos reported significantly more drug using friends and higher levels of general sensation seeki ng compared to White/Other adolescents. Group-Based Trajectory Analyses Group-based trajectory analysis was used to determine the number and shape of trajectories of aggression, delinque ncy, and substance use as indica ted in the first aim of this study. The number of groups identified as most parsimonious and descriptive of the developmental patterns in the data for each outcome were based on maximizing the Bayesian Information Criterion (BIC) score for the outcome of interest. The probability that a particular model was the correct model was based on an estimate developed by Kass and Wasserman (1995). This estimate is calculated by: e BIC j BIC max j e BIC j BIC max where BICmax is the maximum BIC score of the J different models and BICj is the BIC score of a particular model. In cases where th e BIC score did not cleanl y identify a preferred
94 number of groups, model selection was based on domain knowledge and the objectives of the analysis as recommended by Nagin (2005). It is important to reiterate that the number of groups selected for each outcome is not immutable and even individuals who are assi gned to a particular trajectory group do not necessarily follow that groups trajectory in lock step. These trajectory groups are meant to serve as a useful heuristic devi ce in describing developmental patte rns in the data. Each model was estimated for the entire sample and then evaluated via cross-tabulation to determine if there were gender differences in trajectory group membership. Aggression Examinations of the BIC scores resulted in strong evidence in support of a 5 group model for aggression (See Table 5-4). The final traj ectory model of aggr ession across the middle school years is presented in Figure 5-1. Of thos e with complete data, 85.6% were assigned to a trajectory group based on a posterior probability greater than 0.60. The assessment of aggression used in this study is a sum scor e of self-reported engagement in aggressive behaviors over the past month. As such, the numeric value represente d on the y-axis refers to discrete numbers of aggressive acts over the past month. See Table 5-5 for average levels of aggressive behavior within each trajectory group. Ther e was a consistently low group (n = 441, 27% of the sample). This group displayed very low leve ls of aggression across the mi ddle school years. There were two groups with increasing levels of aggressi on across middle school, a slowly increasing group (n = 599, 37% of the sample) and a rapidly increa sing group (n = 344, 21% of the sample). Both of these groups began with simila r levels of aggression which were only slightly higher than the consistently low group in the 6th grade. However, over 7th and 8th grades the slowly increasing group roughly doubled their reports of engagement in aggressive behaviors while the rapidly increasing group roughly tripled their engagement in aggressive behaviors. For the rapidly
95 increasing group, most of the increase in a ggressive behavior occurred between the 6th and 7th grades. Another group showed an interesting patter n of high initial levels of aggression in the 6th grade which increased between 6th and 7th grades followed by a decrease in reported rates of aggression to levels comparable to the consistently low group in the 8th grade. This group was labeled increasing/decreasing (n = 62, 4% of the sample). Finally, a chro nic highly aggressive group of individuals was identified (n = 183, 11% of the sample). These individuals reported chronically high levels of a ggression upon entry into the 6th grade and these reports remained high throughout middle school. Table 5-4. Using BIC to select the number of groups to in clude in the aggression model BIC BIC Probability No. of groups (N = 1629) (N = 4887) correct model 1 -18221.79 -18222.88 0.00 2 -17576.25 -17579.55 0.00 3 -17453.42 -17458.92 0.00 4 -17428.50 -17436.19 0.00 5 -17354.71 -17364.60 1.00 6 -17369.50 -17381.59 0.00 Note: N = 1629 refers to the number of participants; N = 4887 refers to the total number of assessments across the three years of the study
96 0 5 10 15 20 25 30 35 40 6th7th8th GradeYSR Low Low increasing High increasing Increasing/decreasing High Figure 5-1. Five group aggr ession trajectory model Table 5-5. Descriptive statistics within aggression trajectory group Aggression Trajectory Group Low Slow Increasing Rapid Increasing Increasing / Decreasing High Total sample Demographic Variables Gender (% female)** 55% 54% 59% 34% 56% 50% % African American*** 44% 49% 54% 61% 66% 48% % Latino** 32% 29% 24% 16% 21% 30% % Living with Two Parents** 64% 57% 62% 58% 48% 57% School Type (% public) 87% 86% 87% 86% 93% 90% Aggression Variables M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) 6th Grade 5.24 (5.33) 9.98 (6.88) 12.98 (6.60) 23.60 (9.60) 32.30 (5.41) 12.54 (10.54) 7th Grade 6.29 (5.15) 15.42 (7.24) 29.83 (6.33) 30.00 (7.43) 32.64 (7.30) 18.98 (12.08) 8th Grade 6.23 (4.53) 20.58 (7.11) 32.27 (6.25) 6.16 (6.11) 33.99 (6.60) 19.39 (12.42) Note : p < .05; ** p < .01; *** p < .001
97 An analysis of the numbers of males and fe males in each trajectory group found significant gender differences in only one group, 2(4, N = 1,629) = 13.67, p = .008. The increasing/decreasing aggression gr oup was comprised of significantly more males (n = 41) than females (n = 21). The number of males and fema les were not significantly different within the other four trajectory groups (S ee Table 5-5). In addition, ther e were significant differences found for race/ethnicity and household struct ure among individuals following different trajectories of aggressive beha vior during middle school. As s hown in Table 5-5, there were significantly more African American adolescents, 2(4, N = 1,629) = 29.65, p < .001, who followed a trajectory of increasing/decreasing or high aggression and sign ificantly fewer Latinos, 2(4, N = 1,629) = 15.46, p = .004, following those same trajectories. Furthermore, adolescents in the increasing/decreasing and high trajectory groups were less likely to live in a two parent household in the 6th grade, 2(4, N = 1,625) = 15.86, p = .003. Delinquency Examination of the BIC scores did not produ ce conclusive results regarding the optimal number of delinquency trajectory groups. As disc ussed previously, for some constructs the BIC score cannot be used to determine the best fitti ng model because the score continues to rise with the addition of more groups. As such, a 6 group model for the deli nquency construct was selected as the best fitting model based on ma ximizing parsimony without sacrificing meaningful variation in developmental trajectories. Models with fewer groups did not fully capture the rich trends in the data while models with more groups did not conti nue to add any new meaningful developmental patterns of delinque ncy across middle school. The final trajectory model of delinquency acr oss the middle school years is presented in Figure 5-2. Of those with complete data, 90.2% were assigned to a trajectory group based on a posterior probability greater than 0.60. Self-reported delinquent acts over the past year were
98 summed for each adolescent and as such, the numeric values represented on the y-axis refer to individual acts of delinquency over the past year. Table 5-6 presents average levels of delinquent behavior within each trajectory group and for the sample overall. The trajectories observed for delinquency across middle school were similar to those observed for middle school aggres sion. There were two consistently low groups, one reported virtually no delinquency across middle school (n = 714, 44% of the sample) while the other reported consistently low levels of delinquency (n = 227, 14% of the sample). Notably, these 2 groups comprise 58% of the sample, indicating that most youth engage in no or very few delinquent acts during early adolescence. There were also two groups with increasing levels of delinquency across middle school. One increased more slowly (n = 302, 19% of the sample) while the other evinced rapid increases in delinquen cy (n = 140, 9% of the sample). The slowly increasing delinquency group began with initial levels of delinquency near zero in the 6th grade which subsequently increased acro ss middle school to moderate levels. The rapidly increasing group began with slightly higher levels of delinqu ency than the slowly in creasing group in the 6th grade which increased to levels comparable to the chronically high delinquent group by 8th grade. A group similar to the increasing/decr easing aggression group was also identified for delinquency. However, this group reported more stable to slightly increasing delinquency between 6th and 7th grades followed by a decrease to levels slightly higher than the low group in 8th grade. This group was labeled stable/decreasi ng (n = 179, 11% of the sample). Finally, a chronic highly delinquent group of individuals were iden tified (n = 73, 5% of the sample). These individuals reported chronically high levels of delinquency upon en try into the 6th grade which remained high with slight in creases throughout middle school.
99 0 5 10 15 20 25 30 6th7th8th GradeSum Score None Low Low increasing High increasing Increasing/decreasing High Figure 5-2. Six group deli nquency trajectory model Table 5-6. Descriptive statistics within delinquency trajectory group Delinquency Trajectory Group None Low Slow Increasing Rapid Increasing Stable / Decreasing High Total sample Demographic Variables Gender (% female)*** 60% 53% 55% 48% 46% 37% 50% % African American*** 46% 51% 48% 61% 63% 62% 48% % Latino* 29% 33% 27% 24% 19% 19% 30% % Living with Two Parents 61% 58% 64% 54% 56% 52% 57% School Type (% public) 87% 84% 87% 90% 89% 95% 90% Delinquency Variables M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) 6th Grade 0.82 (1.20) 4.33 (2.29) 1.92 (1.62) 4.86 (3.28) 9.72 (6.65) 17.95 (8.60) 3.83 (5.55) 7th Grade 1.20 (1.46) 4.28 (2.22) 6.50 (3.70) 14.15 (6.25) 12.05 (6.65) 22.09 (9.63) 6.16 (7.37) 8th Grade 1.72 (2.05) 3.91 (2.23) 9.80 (4.78) 21.50 (7.77) 6.58 (4.49) 25.39 (10.12) 6. 64 (8.17) Note : p < .05; ** p < .01; *** p < .001
100 There were significant gender and racial /ethnic differences found for delinquency trajectory group membership. As can be seen in Table 5-6, there was a slightly higher percentage of females in the 3 lower deli nquency trajectory groups. However, the high delinquency group showed significantly fewer females than all of the other groups, 2(5, N = 1,635) = 27.07, p < .001. In addition, there were si gnificantly more African American adolescents who followed a trajectory of ra pidly increasing, stable /decreasing, or high delinquency, 2(5, N = 1,635) = 27.42, p < .001. There were significantly fewer Latinos following trajectories of stable/decreasing and high delinquency, 2(5, N = 1,635) = 13.74, p = .017. Substance Use Examination of the BIC scores resulted in strong evidence in support of a 5 group model of substance use across middle school (See Table 57). The five group model of substance use during middle school is presente d in Figure 5-3. Of those w ith complete data, 94.6% were assigned to a trajectory group ba sed on a posterior probability greater than 0.60. Adolescents self-reported use of tobacco, alcoho l, marijuana, inhalants, and ot her hard drugs. Higher values on this sum score of substance use represent more frequent use of multiple substances (See methods section for more detailed description). Values higher than four on the y-axis can only be obtained via use of multiple substances. See Table 5-8 for average levels of substance use within each trajectory group.
101 Table 5-7. Using BIC to select the number of groups to incl ude in the subs tance use model BIC BIC Probability No. of groups ( N = 1586) ( N = 4758) correct model 1 -8041.25 -8041.80 0.00 2 -5403.66 -5406.96 0.00 3 -5081.37 -5087.42 0.00 4 -5045.54 -5054.33 0.00 5 -5020.91 -5032.45 1.00 6 -5039.33 -5053.62 0.00 Note : N = 1586 refers to the number of participants; N = 4758 refers to the total number of assessments acr oss the three years of the study 0 2 4 6 8 10 6th 7th 8th GradeSum Score None Very low Low Low increasing High increasing Figure 5-3. Five group subs tance use trajectory model Most notably, the majority of adolescents re port no use of substances during the middle school years (n = 798, 50% of the sample) or very little use (n = 552, 35% of the sample). This leaves 15% of individuals w ho do become more involved with substance use during middle school which places them at great risk for fu ture negative developmental outcomes in late
102 adolescence and throughout adul thood. Of these remaining individuals, one group reported relatively stable low levels of substance use (n = 117, 7% of the sample). Finally, there are two groups which report increasing levels of substan ce use across middle school, a slowly increasing group (n = 99, 6% of the sample) and a rapidly in creasing group (n = 20, 1% of the sample). The slowly increasing trajectory group be gan with initial levels of subs tance use near zero in the 6th grade. By 7th grade, substance use in this group has in creased to levels comparable to the low level substance use trajectory group. Subsequently, levels of subs tance use continued to escalate such that by 8th grade, levels of substance use were si milar to those of the rapidly increasing trajectory group. Most of the increase in substance use occurred between 7th and 8th grade for the slowly increasing group. On the contrary, most of the increase in substance use occurred between the 6th and 7th grades for the rapidly increasing s ubstance use trajectory group. This group began with the highest levels of substance use in the 6th grade and quickly tripled in their reports of substance use by 7th grade with maintenance of high level usage in the 8th grade. While this is a small subgroup of individuals, they represent those adolescents at most risk for more serious problems. Interestingly, there we re no demographic differences in substance use trajectory group membership; however, there were small sample sizes in 3 of the groups.
103 Table 5-8. Descriptive Statistics wi thin Substance Use Trajectory Group Substance Use Trajectory Group None Very Low Low Slow Increasing Rapid Increasing Total Sample Demographic Variables Gender (% female) 55% 55% 48% 60% 50% 50% % African American 51% 53% 48% 48% 55% 48% % Latino 26% 28% 31% 25% 30% 30% % Living with Two Parents 59% 57% 58% 69% 60% 57% School Type (% public) 88% 86% 85% 90% 90% 90% Substance Use Variables M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) M ( SD ) 6th Grade 0 (0) 0.43 (0.69) 2.06 (1.85) 0.39 (0.78) 3.44 (3.97) 0.50 (1.52) 7th Grade 0 (0) 0.76 (0.87) 3.62 (2.32) 2.14 (2.04) 10.30 (4.92) 0.93 (2.10) 8th Grade 0 (0) 1.31 (1.30) 2.48 (1.73) 8.39 (4.06) 9.21 (5.63) 1.30 (2.74) Note : ** p < .01; *** p < .001 Dual Trajectory Analysis Dual trajectory analysis is an advanced form of group-based trajectory modeling where trajectory models are simultaneously estimated for two separate outcomes of interest (e.g., aggression and substance use) along with the probab ility that individuals who follow a particular trajectory on one of the outcomes (e.g., aggression ) also follows a particular trajectory on the other outcome (e.g., substance use). These an alyses provide information on the temporal associations between drug use a nd aggression and delinquency as indicated in the first aim of the study.
104 As discussed previously, it is well established that increases in aggressive behavior precede increases in delinquent behavi or during middle school, both of which precede the onset of substance use. As such the current study evaluate d the probability of memb ership in a particular delinquency or substance use trajectory given pr evious assignment to a particular aggression trajectory. The probability of membership in a particular substance use trajectory given assignment to a particular delinquency trajectory group was also evaluated. Aggression and Delinquency Based on the models established in the singl e outcome group-based trajectory analyses, a dual trajectory analysis was estimated for a five group aggression model and a 6 group delinquency model. The probabilities of following a particular trajectory of delinquency given membership in a particular aggression tr ajectory group are presen ted in Table 5-9. Table 5-9. Probability of deli nquency trajectory group membership given aggression trajectory group membership Delinquency Trajectory Group None Low Slow Increasing Rapid Increasing Stable / DecreasingHigh Aggression Trajectory Grou p Low [88% 11%] 0% 0% 0% 0% Slow Increasing 6% [68% 24%] 0% 2% 0% Rapid Increasing 1% 11% [54% 30%] 3% 0% Increasing / Decreasing 2% 17% 0% 0% [72%] 9% High 0% 0% 0% [26% 19% 55%] A noticeable pattern between trajectories of aggression and delinquency emerged such that individuals following a particular trajectory of aggression tended to follow a similar pattern of delinquency during middle school. For example, individuals in the low aggression trajectory
105 group have a very high probability of also followi ng a trajectory of no delinquency to low levels of delinquency. The high aggres sion trajectory group was most st rongly associated with rapidly increasing or high trajectories of delinquency. In addition, a couple of intere sting patterns emerged. The slowly increasing aggression group had the highest probability of following a lo w trajectory of delinquency. Similarly, the rapidly increasing aggression group had the highest probability of following a slowly increasing delinquency trajectory. This il lustrates the developmental trend of aggressive behavior preceding delinquent behavior while emphasizing the association between increases in both. Another interesting trend invol ves the pattern of probabilitie s of delinquency trajectory group membership associated with the increasin g/decreasing aggression tr ajectory group. The vast majority of individuals who follow a trajectory of increasing/de creasing aggression (72%) are likely to follow a similar trajectory of stable/decreasing delinquency across middle school. However, there is an interesting spl it among the remaining individuals in the increasing/decreasing aggression group illustrated by a 19% probability of following a trajectory of low to no delinquency and a 9% probability of following a trajectory of high delinquency. These findings raise many questions regarding why there is such diversity in the probability of belonging to a particular deli nquency trajectory group given a pa ttern of increasing/decreasing aggression during middle school. Of note, is the finding that the increasin g/decreasing trajectory is the only aggression trajectory group which has significa ntly more males than females. Aggression and Substance Use Dual trajectory analysis was used to estimate the probability that individuals following a particular trajectory of aggression would also foll ow a particular trajector y of substance use for a five group aggression model and a 5 group substa nce use model. These probabilities are presented in Table 5-10.
106 Table 5-10. Probability of s ubstance use trajectory group membership given aggression trajectory group membership Substance Use Trajectory Group None Very Low Low Slow Increasing Rapid Increasing Aggression Trajectory Group Low [91%] 0% 8% 1% 0% Slow Increasing [58% 30%] 6% 5% 1% Rapid Increasing [35% 37%] 13% 10% 4% Increasing / Decreasing [25% 14% 61%]0% 0% High [28% 10% 36% 11% 15%] A similar pattern emerged between trajectori es of aggression and substance use as was observed between trajectories of aggression and delinquency, although somewhat muted given the lower levels of reported s ubstance use during the middle school years. Individuals following a trajectory of low aggression had a high proba bility of abstaining from substance use during middle school. Among the other trajectories of a ggression, more problematic trajectories had a high probability of also following more severe substance use trajectories. The high aggression trajectory group was at most risk for following a trajectory of stable low or increasing substance use, both slow and rapid. The increasing/decreasing aggression traject ory group had a 61% probability of also following a trajectory of stable low substa nce use. This is noteworthy because the increasing/decreasing aggressi on trajectory group has more males than females, a high probability of following a trajecto ry of stable/decreasing deli nquency, and has a high probability of stable low substance use. These individuals seem to repres ent a group of adolescents who
107 have a short period of increased problem behavior around 7th grade accompanied by low level substance use experimentation. Further evaluation of the developmental history, personality characteristics, and social environment of these students will provide a mo re complete picture of the antecedents of this pattern of behavior. In addition, continued evaluations of engagement in problem behaviors throughout high school and young adulthood will determine the developmental outcomes associated with th ese particular middle school trajectories. Delinquency and Substance Use A dual trajectory analysis estimated the pr obability of membership in a particular substance use trajectory group gi ven a particular trajectory of delinquency across middle school based on the previously established 6 group de linquency model and 5 group substance use model (See Table 5-11). Table 5-11. Probability of s ubstance use trajectory group membership given delinquency trajectory group membership Substance Use Trajectory Group None Very Low Low Slow Increasing Rapid Increasing Delinquency Trajectory Group None [86% 14%] 0% 0% 0% Low [60% 33%] 4% 2% 1% Slow Increasing [40% 41%] 6% 11% 2% Rapid Increasing [34% 29% 35%] 0% 3% Stable / Decreasing [ 19% 47%] 1% [24% 9%] High [18% 9% 42% 15% 16%]
108 Similar to both of the previous dual trajectory analyses, more problem atic trajectories of delinquency are associated with more severe traj ectories of substance us e during middle school. Trajectories of low and no deli nquency were most strongly asso ciated with no to very low substance use. The slowly increasing delinquency trajectory had slightly higher probabilities of following a low or increasing substance use traj ectory. The rapidly increasing delinquency trajectory had a higher probability of following a trajectory of stable low substance use but a lower probability of following a trajectory of sl ow increasing substance use compared to the slow increasing delinquency group. The high delinquency trajectory group was at most risk for following a trajectory of stable low or increas ing substance use, both slow and rapid. Similar to aggression, the stab le/decreasing delinquency trajec tory had an unusual pattern of association with the substan ce use trajectories. The stable /decreasing delinquency trajectory had a high probability of either following a trajectory of no to very low substance use or a trajectory of increasing substance use, both slow and rapid. This suggests two different explanations for the association between these two trajectories. It may be possible that there is one group of adolescents who are desisting from problem behavior in general while another group of adolescents may be transitioning from e ngagement in delinquent behavior to increased substance use. As indicated, gender differences in group me mbership only appear in the high delinquency trajectory group. This group has more males than females in addition to a high probability of following a stable low or increasing substance us e trajectory. Furthermore, gender differences are observed for the increasing/decreasing aggres sion trajectory group as well which is also strongly associated with the st able low substance use trajecto ry. Yet, there are no observed
109 gender differences in any of the substance use trajectories. This may be due, in part, to the low reports of substance use overall among this middle school sample. Mediation Analyses The mediating role of aggression and delinque ncy (Aim 2) on the association between sensation seeking, delinquent peer s, and the interaction of sensa tion seeking with deviant peers on substance use was evaluated via HLM. An in tercepts as outcomes model was estimated for the mediation analyses. As such, only level one (w ithin person) predictors were included in the model. All level one predictors were person-center ed to test the influence of changes within an individual on their subsequent behavior. Gender, race/ethnicit y, household structure, and school type were controlled for in all mediation analyses. The role of aggression and delinquency as medi ators of the interaction between sensation seeking and delinquent peers on substance us e was evaluated (See Figure 5-6). Follow up evaluations of the main effects of sensation s eeking and delinquent peers were evaluated only if the two-way interaction between sensation seek ing and deviant peers was not statistically associated with substance use. Figure 5-6. Mediation model evaluating aggression as the mediator Figure 5-6 illustrates both the di rect association between the independent variable and the dependent variable (path c ) and the indirect association between the independent variable and the Aggression or Delinquency (Mediator) a b Sensation Seeking X Friend Deviancy Substance Use (Independent Variable) c (Dependent Variable)
110 dependent variable throu gh the mediator (paths a and b). The indirect effect was estimated by multiplying the coefficients for path a and path b and the significance of the indirect effect was approximated through the use of Sobel tests. An evaluation of path c revealed that there was not a sign ificant direct asso ciation of a twoway sensation seeking by friend delinquency intera ction on substance use. Due to the lack of association between the interact ion and substance use, followup analyses of associations between the main effects of person-centered se nsation seeking and pers on-centered delinquent peer association on substance use were evaluated. Aggression and Delinquency as Med iators of Sensation Seeking An initial evaluation of the association be tween person-centered sensation seeking and substance use was statistically significant (path c), = 0.23, SE = 0.04, t (5039) = 6.04, p < .001. Experiencing levels of sensation seeking that were higher than average was associated with greater reported substance use. Separate eval uations of the main effect of person-centered sensation seeking on both aggression = 2.89, SE = 0.19, t (5095) = 15.49, p < .001 and delinquency = 1.16, SE = 0.11, t (5098) = 10.64, p < .001 (path a) were also significant. The direction of effect was similar to substance use, such that increases in sensation seeking within an individual were associated with incr eases in both aggression and delinquency. The final evaluation of aggre ssion and delinquency as mediator s of the association between sensation seeking and substance use were evaluated separately for each potential mediator. First, both person-centered aggression and person-cent ered sensation seeking were simultaneously regressed onto substance use. The previously significant association be tween sensation seeking and substance use remained statistically significa nt; however, this effect was partially mediated by individual changes in aggression. Sobel tests evaluated the strength of the indi rect effect and found it to be statistically significant ( z statistic = 8.36, p < .001).
111 Second, the mediating influence of delinque ncy on the association between sensation seeking and substance use was evaluated. Pe rson-centered delinquency and person-centered sensation seeking were both simultaneously regr essed onto substance use. The previously significant association between sensation seeking and substan ce use remained statistically significant; however, this effect was partially mediated by individual ch anges in delinquency ( z statistic = 8.19, p < .001). As such, increases in substance use that are associated with increases in sensation seeking are partially accounted for by individual increases in both aggression and delinquency. Aggression and Delinquency as Mediators of Friend Delinquency The main effect of person-centered friend de linquency on substance use was also found to be statistically significant (path c), = 0.09, SE = 0.01, t (6722) = 10.64, p < .001. Associating with more delinquent friends than typical was associated with gr eater reported substance use. Person-centered friend delinquency was also significantly associated with both aggression = 0.94, SE = 0.03, t (6805) = 31.03, p < .001and delinquency = 0.58, SE = 0.02, t (6808) = 25.28, p < .001 (path a). Increases in association with delinquent friends were associated with increases in aggression as well as delinquency. First, aggression was evaluated as a medi ator of the association between friend delinquency and substance use. Both personcentered aggression and person-centered friend delinquency were simultaneously regressed onto substance use. The previously significant association between friend delinquency and subs tance use remained st atistically significant; however, this effect was partially mediated by in dividual changes in aggre ssion. The strength of the indirect effect was determined to be statistically significant via a Sobel test ( z statistic = 7.29, p < .001).
112 Second, delinquency was evaluated as a mediator of the association between friend delinquency and substance use. Person-cente red delinquency and pe rson-centered friend delinquency were both simultaneously regressed onto substance use. Individual changes in delinquency were found to partially mediate th e association between friend delinquency and substance use. Sobel tests revealed that the indirect effect was stat istically significant ( z statistic = 9.45, p < .001). These results indicate that indi vidual changes in aggr ession and delinquency both partially account for th e association between subs tance use and delinquent peer association. In sum, increases in both sensation seek ing and delinquent peers were significantly associated with greater substance use, aggression, and delinquency. In addition, increases in aggressive behavior partially mediated the asso ciation of both sensati on seeking and delinquent peers with substance use. Increases in delinquent behavior partially medi ated the association of both sensation seeking and delinquent peers with substance use. The Moderating Role of Gender The moderating role of gender on the associ ation between sensati on seeking, delinquent peers, and the interaction of sensation seeki ng with delinquent peers (Aim 3) on aggression, delinquency, and substance use were evaluated vi a HLM. Evaluations of gender as a moderator of the interaction betwee n sensation seeking and deviant peer s on the outcomes of interest were evaluated. If the three-way cross-level intera ction between gender, se nsation seeking, and delinquent peers was not statisti cally significant, follow up eval uations of the two-way cross level interaction betwee n gender and sensation seeking as well as between gender and delinquent friends were evaluated. As disc ussed previously, all predictors were person-mean centered to provide information about how indi vidual changes in thes e predictors influen ced the associations of interest. Household structur e, race/ethnicity, and school type were controlled for in all analyses.
113 The evaluation of a three way interaction be tween person-mean cente red sensation seeking, person-mean centered friend delinquency, and gender was not signifi cant for aggression, delinquency or substance use. Two-way cross le vel interactions were evaluated between gender and person-mean centered sensation seeking as well as gender and person-mean centered friend delinquency for all three outcomes of interest. Sensation Seeking by Gender A significant interaction between gender and person-mean centered se nsation seeking was found for substance use (See Table 5-12). The association between i ndividual changes in sensation seeking and aggressive or delinquent behavior was not moderated by gender. Table 5-12. Hierarchical linear models examin ing the association between gender, sensation seeking, and substance use Parameter Model 1 Model 2 Model 3 Model 4 Fixed Effects Intercept 00 0.85*** (0.03) 0.91*** (0.13) 0.93*** (0.14) 0.93*** (0.14) Latino 01 -0.02 (0.09) 0.02 (0.11) 0.02 (0.11) Black 02 -0.10 (0.08) -0.16 (0.10) -0.16 (0.10) Gender 03 0.04 (0.06) 0.05 (0.07) 0.05 (0.07) School Type 04 -0.01 (0.09) 0.02 (0.10) 0.02 (0.10) Household Structure 10 -0.03 (0.07) -0.06 (0.08) -0.06 (0.08) Sensation Seeking 20 0.23*** (0.04) 0.32*** (0.06) Gender X Sensation Seeking 21 -0.18* (0.08) Variance Components Level 1 Within-person residual ei j 3.37*** (1.83) 3.37*** (1.83) 3.25*** (1.80) 3.25*** (1.80) Level 2 Intercept 0 j 1.10 (1.05) 1.11 (1.05) 1.30 (1.14) 1.30 (1.14) Pseudo R2 and Goodness of Fit R2 (within person variance) 0.00 0.04 0.04 R20 (intercept variance) 0.00 0.00 0.00 Deviance (-2LL) 29,219.99 29,232.20 21,723.70 21,722.87 Deviance (-2LL) -12.21 7508.50 0.83 Note : p < .05, ** p <.01, *** p < .001
114 A follow-up evaluation of how individual change s in sensation seeking impacted substance use by gender revealed a more pronounced effect for females. Both males and females reported lower rates of substance use when their own indi vidual levels of sensation seeking were lower than average and higher rates of substance use wh en their individual levels of sensation seeking were higher than average. However, higher levels of sensation s eeking showed a more pronounced influence on females substance us e compared with male s (See Figure 5-4). Figure 5-4. Gender moderating the association between person-centered sensation seeking and substance use Friend Delinquency by Gender A significant interaction between gender and person-mean centered friend delinquency was found for delinquent behavior only (See Table 5-13). The association between individual changes in friend delinquency and aggressive be havior or substance us e was not moderated by gender.
115 Table 5-13. Hierarchical linear models exam ining the association between gender, friend delinquency, and delinquency Parameter Model 1 Model 2 Model 3 Model 4 Fixed Effects Intercept 00 5.27*** (0.11) 3.44*** (0.38) 3.42*** (0.38) 3.42*** (0.38) Latino 01 -0.18 (0.29) -0.19 (0.29) -0.19 (0.29) Black 02 1.17*** (0.28) 1.17*** (0.28) 1.17*** (0.28) Gender 03 1.88*** (0.21) 1.92*** (0.21) 1.92*** (0.21) School Type 04 0.51 (0.29) 0.51 (0.29) 0.51 (0.29) Household Structure 10 -0.23 (0.24) -0.22 (0.24) -0.22 (0.24) Friend Delinquency 20 0.58*** (0.02) 0.52*** (0.03) Gender X Friend Delinquency 21 0.11* (0.05) Variance Components Level 1 Within-person residual ei j 28.26*** (5.32) 28.21*** (5.31) 20.06*** (4.48) 19.99*** (4.47) Level 2 Intercept 0 j 20.89 (4.57) 19.73 (4.44) 23.03 (4.80) 23.06 (4.80) Pseudo R2 and Goodness of Fit R2 (within person variance) 0.00 0.29 0.29 R20 (intercept variance) 0.06 0.00 0.00 Deviance (-2LL) 45,435.97 45,328.08 43,457.16 43,445.70 Deviance (-2LL) 107.89 1870.92 11.46 Note : p < .05, ** p <.01, *** p < .001 A follow-up evaluation of how individual change s in association with delinquent friends impacted delinquent behavior by gender revealed a stronger influence of delinquent friends on females compared to males. Both males and fema les reported lower rates of delinquent behavior when associating with fewer delinquent peers th an they do on average. Both genders also reported increases in reports of delinquent behavior when associ ating with a greater number of delinquent peers than their individual average (See Figure 5-5). Figure 5-5 illustrates the difference between the genders is not in the direction of the effect but rather in the strength of the effect. Females show a much more pronounced in fluence of changes in their delinquent peer association on their delinquent behavior.
116 To summarize, females reported a more pr onounced negative impact associated with increases in sensation seeki ng and delinquent peer association compared with males. Specifically, females reported great er increases in delinquent behavi or associated with individual increases in delinquent peer asso ciation and greater increases in substance use associated with individual increases in sens ation seeking compared with males. There were no gender differences in the association between individual increases in sensati on seeking and delinquent peers on aggressive behavior. Figure 5-5. Gender moderating the association between person-centered friend delinquency and delinquency The Moderating Role of Race / Ethnicity The moderating role of race/ethnicity (Aim 4) on associations between antisocial behaviors (aggression, delinquency, and substance use), sensation seeking, and peer delinquency was evaluated via HLM. These analyses determined if there were differences in associations between the predictors and the outcomes of interest among Black, Latino, and White/Other adolescents.
117 Evaluations of race/ethnicity as a moderator of the interaction between sensation seeking and deviant peers on the outcomes of in terest were evaluated. If the three-way cross-level interaction between race/ethnicity, sensation seeking, and de viant peers was not stat istically significant, follow up evaluations of the two-way cross level interaction between race/ ethnicity and sensation seeking as well as between race/ethnicity and deviant peers were evaluated. As discussed previously, all predictors were group-mean centered to provide information about how individual changes in these predictors influenced the associ ations of interest. Gender, household structure, and school type were contro lled for in all analyses. Evaluations of a three way interaction between race/ethnicity person-centered sensation seeking, and person-centered fr iend delinquency were evaluated for aggression, delinquency, and substance use. A significant Latino X sensati on seeking X friend delinquency interaction was found for aggression (See Table 5-14). Table 5-14. Hierarchical linear models examini ng the association between Latino race/ethnicity, sensation seeking, friend delinquency, and aggression Parameter Model 1 Model 2 Model 3 Model 4 Fixed Effects Intercept 00 16.20*** (0.18) 15.21*** (0.68) 15.62*** (0.71) 15.62*** (0.71) Latino 01 -1.33** (0.51) -1.36* (0.56) -1.36* (0.56) Black 02 2.41*** (0.48) 2.09*** (0.52) 2.09*** (0.52) Gender 03 0.72* (0.36) 0.48 (0.41) 0.48 (0.41) School Type 04 -0.18 (0.57) 0.61 (0.58) 0.61 (0.58) Household Structure 10 0.09 (0.41) -0.40 (0.44) -0.40 (0.44) Friend Delinquency 20 0.87*** (0.04) 1.00*** (0.09) Friend Delinquency X Latino 21 -0.14 (0.12) Friend Delinquency X Black 22 -0.16 (0.11)
118 Table 5-14. Continued. Parameter Model 1 Model 2 Model 3 Model 4 Fixed Effects Sensation Seeking 30 2.11*** (0.17) 1.91*** (0.34) Sensation Seeking X Latino 31 0.54 (0.44) Sensation Seeking X Black 32 0.03 (0.43) Friend Delinquency X Sensation Seeking 40 -0.05* (0.02) -0.13*** (0.05) Friend Delinquency X Sensation Seeking X Latino 41 0.16* (0.07) Friend Delinquency X Sensation Seeking X Black 42 0.08 (0.06) Variance Components Level 1 Within-person residual ei j 85.65*** (9.25) 85.49*** (9.25) 60.93*** (7.81) 60.81*** (7.80) Level 2 Intercept 0 j 58.49 (7.65) 56.17 (7.49) 68.96 (8.30) 69.03 (8.31) Pseudo R2 and Goodness of Fit R2 (within person variance) 0.00 0.29 0.29 R20 (intercept variance) 0.04 0.00 0.00 Deviance (-2LL) 52,823.44 52,736.02 38,277.89 38,282.19 Deviance (-2LL) 87.42 14458.13 -4.30 Note : p < .05, ** p <.01, *** p < .001 Follow-up analyses revealed th at increases in both sensati on seeking and delinquent peers were associated with increases in aggression regardless of race/ethnicity. However, Latinos aggressive behavior was more st rongly influenced by individual changes in sensation seeking compared to other race/ethnicities. In addition, other race/ethnicities repo rted a greater influence of individual changes in delinquent peer association on aggressive behavior compared to Latinos (See Figure 5-7). There were no significant three way interactions between race/ethnicity, sensation seeking, and friend delinquency on either delinquency or substance use.
119 Note: Low SS refers to individual s who report lower sensation seeking; High SS refers to individuals who report hi gher sensation seeking. Figure 5-7. A three-way inter action between Latino race/ethnicity, sensation seeking, and delinquent peer association on aggressive behavior Given the lack of significant three way interactions between race/ethnicity, sensation seeking, and friend delinquency on either delinquen cy or substance use, follow-up evaluations of two way interactions were conducted. Evaluations of the interaction betw een race/ethnicity and person-centered sensation seeking were evaluate d for both delinquency and substance use. None of these two way interactions were statistically significant. Evaluations of the moderating effect of race/ethnicity on the asso ciation between person-centered friend delinquency and both substance use and delinquency also did not reveal significant results. As such, race/ethnicity did not moderate the associations be tween individual changes in sens ation seeking or delinquent peer association on delinquency or substance use.
120 To summarize, race/ethnicity moderated the interaction between sensation seeking and delinquent peers on aggression such that Latinos were somewhat mo re influenced by increases in sensation seeking while other r ace/ethnicities were somewhat mo re influenced by increases in delinquent peer association. Howeve r, despite statistical significan ce, this effect was quite small in size. Given the overall lack of moderation by race/ethnicity in this study, the aforementioned finding for Latinos should be interpreted with caution.
121 CHAPTER 6 DISCUSSION The purpose of this study was to evalua te changes in aggression, delinquency, and substance use across the middle school years and begin to elucidate interconnections between these negative adjustment outcomes as well as their antecedents. Deviant peer association and sensation seeking were evaluated in this study as they each represent cons tructs, one social and the other personality, which are strongly associ ated with increases in a variety of problem behaviors during the middle schoo l years including a ggression, delinquency, and substance use. A key element of this study was an emphasis on evaluating gender differences regarding interconnections between these antisoc ial behaviors and their antecedents. The first aim was to evaluate trajectories of change across middle school for aggression, delinquency, and substance use. Group-based trajectory analyses revealed distinct patterns of change for each outcome of interest. The tr ajectory groups which best modeled aggression and delinquency revealed similar patterns of change across middle school. Both outcomes had stable low groups which reported little if any engagement in the antisoc ial behavior. Both aggression and delinquency also reported grou ps of increasing antisocial behavior as well as groups with stable high antisocial behavior. These groups map well onto Moffitts theory of adolescent limited vs. life-course persistent antisocial beha vior (Moffitt et al., 2001). The increasing groups are most similar to Moffitts adolescent limited offenders. Rates of antisocial behavior are near zero in the 6th grade with steady increases to levels co mparable to the stable high group by the 8th grade. The stable high groups begin 6th grade with high levels of both aggression and delinquency and maintain these levels across middl e school which is similar to what would be expected from a chronic offender. Likewise, th e dual trajectory analysis between aggression and delinquency supports the associa tion between these groups. This analysis revealed a high
122 probability that individuals following increasi ng aggression trajectories will also follow trajectories of increasin g delinquency. In addition individu als in the stable high aggression trajectory had the highest probabi lity of following a trajectory of stable high delinquency. Broidy and colleagues (2003) evaluated multiple sa mples from within and outside of the United States for gender differences in associations between trajector ies of childhood overt aggression and adolescent delinquency. Similar to the re sults of this study, they found significant associations between overt aggression in chil dhood and adolescent delinquency across samples for males. However, they found less support fo r this association among females with the one exception being a sample of females from within th e United States. As such, the findings of this study do not contradict those of Broidy and colleagues (2003) but th ey do highlight the need for more comprehensive evaluations of gender differences in associations between antisocial behaviors from childhood through young adu lthood among individuals from diverse backgrounds and across a wide age range spanning childhood th rough adolescence. The identification of an increasing/decr easing aggression trajectory group and a stable/decreasing delinquency group was unexpected based on previous research. However, the ability to identify trends in the data which would not necessarily have been predicted a priori is one of the key advantages of group-based traj ectory analysis. As such, evaluating the increasing/decreasing aggression trajectory group has the poten tial to provide new insights regarding developmental changes in aggressive behavior and associatio ns with other problem behaviors in middle school. As shown in the dual trajectory analysis between aggression and delinquency, there was a very high (72%) probabili ty of individuals following a trajectory of increasing/decreasing aggre ssion to also follow a trajectory of stable/decreasing delinquency. Of course, results of this evaluation are more e xploratory and could eith er represent a real
123 developmental trend or a statistical anomaly. Hen ce, these results should be considered stepping stones which future studies can use to gain a richer perspective on patterns of change in the development of aggression and delinquency over time. Trajectories of substance use revealed sm all but significant groups of individuals who began to increase in substance use markedly during middle school. These individuals represent adolescents at the highest ri sk of developing substance use problems and other negative adjustment outcomes. Dual trajectory analyses revealed that patterns of high stable aggression and delinquency were strongly associated with increases in substance use as well. These individuals do appear to be at highest risk for con tinuing to follow a pattern of life-course persistent offending. The increasing/decreasing aggression trajector y group had a high probability of consistent but low levels of substance use during middle school while th e stable/decreasing delinquency group had a fairly high probability of showing in creases in substance use during middle school. This particular pattern is interesting because it suggests that there may be a subgroup of adolescents for which initiation of substance use is accompanied by decreases in aggressive or delinquent behavior. It is possible that there is a small but significant subgroup of adolescents who desist in aggressive and delinquent behavior after transitioning to substance use. Again, given that both the increasing/ decreasing aggression trajectory group and the stable/decreasing delinquency trajectory group are more exploratory, further research is needed to determine if these patterns of behavior hold across high school and adulthood. Future research should evaluate the possibility that for some indivi duals low levels of substance use may curb aggressive or delinquent tendencies. Unders tanding why this transition would take place and how this association may change across la ter adolescence and adulthood would provide
124 important insights for understand ing the etiology of substance use problems and associations with other antisocial behaviors. Notably, there were very few gender differen ces in trajectory group membership. Only two groups, the increasing/decreasing aggressi on group and the stable high delinquency group, consisted of significantly more males compared to females and there were no gender differences found regarding substance use trajectory group memb ership. The lack of gender differences in trajectories of aggression, delinquency, and substance use imp lies that the developmental progression of engagement in these behaviors during middle sc hool is roughly equivalent for both males and females. This finding is in line with previous research on gender differences in patterns of change for childhood aggressive be havior (Broidy et al., 2003). Broidy and colleagues evaluated trajectories of childhood overt aggressive beha vior in samples of children from the United States as well as internationall y. There were very few differences between males and females in patterns of change in overt aggression across ch ildhood and into early adolescence. However, a lack of gender differe nces in patterns of cha nge does not necessarily imply that average rates of these behaviors are the same for both males and females. In fact, this study found that males reported somewhat highe r average rates of aggression in the 6th grade and significantly higher average ra tes of delinquency throughout mi ddle school. These results support the idea that males and females follow si milar patterns of change during middle school regarding aggression, delinquency, and substance us e. The greater number of males who follow a trajectory of stable high delinquency may account for the gender difference in average rates of this behavior. The general lack of gender differences in trajectories of aggression, delinquency, and substance use during middle school has strong imp lications for prevention programming. It is
125 important to emphasize that prevention of aggre ssive, delinquent, and substance using behaviors at this early age is no t just limited to males. Future re search should continue to evaluate underlying mechanisms associated w ith patterns of change in thes e behaviors separate for males and females. This will help determine whic h individual and contextual antecedents of aggression, delinquency, and substance use are salie nt to the developmental progression of these behaviors for males and females. Racial/ethnic differences were found in trajec tory group membership in this sample of urban middle school adolescents for both aggre ssion and delinquency. Black adolescents were more likely to follow trajectories of increasing/decreasing, high increasing, or chronically high aggression and delinquency compared to Latino and White/other adolescents. This finding is in line with previous research which has found higher prevalence rates of aggression and delinquency among minority adolescents compared to White adolescents (CDC, 2005). There were no race/ethnic differences in substance use trajectory member ship despite previous research which has found higher prevalence rates of substance use among Wh ite adolescents and increases among Latinos (CDC, 2005). This may be due, in part, to the young age range of this sample. More distinct patterns of racial/ethnic differences in substance use may appear in late adolescence and early adulthood. In addition, this study evaluated an urban sample. It may be that previously reported raci al/ethnic differences in substa nce use are more pronounced among suburban adolescents. These findings have importa nt implications for prevention research and reduction of problem behaviors among urban minor ity adolescents. Eluc idating the underlying risk and protective factors associ ated with initial trajectories of problem behaviors during the middle school years among ethnic minorities will ai d in the development of ethnically sensitive
126 and appropriate intervention strategies. The ul timate goal is to redu ce ethnic and racial disparities in adjustment outcomes in later adol escence and during the tran sition to adulthood. Two risk factors for problem behavior, sensa tion seeking and deviant peer association, were evaluated in this study. Increases or d ecreases in sensation s eeking and deviant peer association relative to an indivi duals average level on those constructs were evaluated to gain a better understanding of how individu al change affects behavior ra ther than simply looking at trends for the overall sample. The potential mediating influences of individual aggression and delinquency on associations between sensation seeking, devian t peer association and substance use were examined. Aggression and delinquency were ev aluated as mediators of associations with substance use given the well doc umented developmental sequence of the emergence of these behaviors during the middle school years. These analyses found th at individual changes in both sensation seeking and delinquent peer association were predictive of increases in substance use as well as both potential mediators (aggression and delinquency). The association between changes in delinquent peer association on subs tance use was partially mediated by individual changes in both aggression and delinquency. Likewise, the associa tion between individual changes in sensation seeking on substance use wa s partially mediated by individual changes in both aggression and delinquency. These findings high light the potential predictive influences of engagement in one type of problem behavior on subsequent engageme nt in other problem behaviors. Hence, part of the influence of more traditional risk factors for substance use (sensation seeking and deviant p eer association) was accounted fo r by individual engagement in aggressive and delinquent behavi or. As such, interventions ai med at reducing engagement in
127 aggressive behavior in late childhood and minimizing engagement in delinquent behavior in early adolescence may reduce substance use in itiation in early and mid adolescence. The conclusion that interventions aimed at reducing one problem behavior should also result in decreases in other problem behaviors is in line with Problem Behavior Theory (Jessor, 1987; Jessor, 1992) as well as the General Theory of Crime (Gottfredson & Hirschi, 1990). However, problem behavior theory asserts that this generalization of intervention effects from one problem behavior to another is due to overlap in common correlates, not direct influences of one problem behavior on the onset of another pr oblem behavior. The general theory of crime asserts that the generalization of intervention effect s is due to the fact th at all forms of problem behavior emerge as part of an individual tendency towards deviance in general. Without contradicting these two theories, th e results of this investigation s uggest that there may be direct associations among problem behaviors. Th ese direct associations among aggression, delinquency, and substance use, at the least, partially account for associations between common correlates of these behaviors. This investigation also evaluated the moderati ng influences of gender and race/ethnicity on associations between sensation seeking, devian t peer association, aggr ession, delinquency, and substance use. The influence of individual chan ges in deviant peer a ssociation and sensation seeking was similar for both males and females with two exceptions. First, increases in associating with delinquent friends had a sign ificantly stronger influence on increases in delinquent behavior amo ng females compared with males. Second, increases in individual sensation seeking had a signifi cantly stronger influe nce on increases in substance use among females compared with males. These findings cl early identify changes in sensation seeking and deviant peer association as risk s for antisocial behavior for both males and females. However,
128 the results of the moderation analyses suggest that deviant peer association and sensation seeking may have a more pronounced influence on females compared to males regarding engagement in antisocial behaviors during middl e school. As such, these cons tructs may convey a specific vulnerability to females in addition to a general risk for both genders. Interestingly, this study found that individual changes in sensation seeking had a stronger influence on Latinos engagement in aggressive behavior while individual changes in delinquent peer association influenced aggr ession more strongly for other race/ethnicities. This finding highlights an important cultural difference in ri sks associated with engagement in problem behavior by suggesting that Latinos may be less influenced by soci al factors and more influenced by individual factors compared to other race/ethnicities. Importantly, increases in both sensation seeking and delinquent peer asso ciation conferred risk regardle ss of race/ethnicity and the interaction of race/ethnicity, sensation seeking, and friend delinquency did not account for a notable amount of within person variance in aggression. However, this finding does indicate that there may be cultural differences in risk factor s for aggression in ear ly adolescence among urban minority adolescents. Importantly, the aforementioned racial/ethnic di fference was the only significant difference found in associations between the outcomes of interest in this study (aggression, delinquency, and substance use) and the risk factors evaluate d in this study (sensation seeking and deviant peer association). As such, most of the effects of sensation seek ing and deviant peer association were common across both gender and race/ethnicity. The implication of this is that individual changes in sensation seeking and deviant peer as sociation do not fully ex plain ethnic and racial differences or gender differences observed in trajectory group me mbership. This finding is important because it begins to rule out potential mechanisms which under lie observed disparities
129 in aggression and delinquency. Further res earch on other correlat es of aggression and delinquency (i.e., family factors) are needed to identify which risk or protective factors are most influential in contributing to group disparities in aggression, delinquency, and substance use. Once identified, interventions ca n be modified to be more et hnically or gender appropriate. Strengths and Implications of Study Design This study was able to evaluate separate longitudinal trajecto ries of aggression, delinquency, and substance ini tiation across the middle school y ears and associations between these trajectories through the use of novel statistical methodology. Group-based trajectory analysis has the capability to go beyond theory by allowing the da ta to determine if there are different subgroups of individuals within the popul ation who differ in terms of their initial levels of a behavior as well as change s in that behavior across time. The ability to confirm the existence of theorized subgroups of individuals (adolescence limited vs. life-course persistent offenders; Moffitt et al., 2001) as well as identif y subgroups within the population that are not necessarily anticipated by theory is a useful tool for moving the field forward. Another strength of the study is the longitudina l nature of the data across the middle school years, as this is when increases in antisocial be haviors often begin to emerge. However, future research would benefit from evaluations of changes in antisocial behaviors from childhood, through adolescence, and into adulthood to provide the most complete picture of interconnections in the developmen t of these behaviors over time. Perhaps most importantly, this study evaluate d gender and racial/ethnic differences in trajectories of antisocial behavi ors and associations between antisocial behaviors among a group of urban adolescents as they transition thr ough middle school. These results contribute to a growing knowledge base regarding pathways to drug use and delinquency among this group of adolescents.
130 Limitations Despite the aforementioned methodologica l strengths, the present study has some limitations. While the sample char acteristics are considered a stre ngth of the study, the results of this study are limited in that they are not generalizable to the general population. Measurement issues associated with accurately assessing antisocial behaviors as well as correlates of antisocial behaviors will always be a challenge. The current study utilized selfreport only; however, future studies would be nefit from including multiple informants (e.g., parents, teachers) to provide a dditional perspectives on the par ticipants behavior (Achenback, McConaughy, & Howell, 1987; Phares, Compas, & Howe ll, 1989). Validity of self-report in the current study was promoted through the use of the bogus pipeline procedure which has been shown to increase the accuracy of reports of tobacco use (Evans, Hansen, and Mittelmark, 1977) as well as other antisocial behaviors (Tourang eau, Smith, & Rasinski, 1997). While this study was completed with paper and pencil, future research would benefit from use of computer based survey formats, a methodological technique which ha s been shown to increase the validity of self reported antisocial behaviors (Booth-Kewley, La rson, & Miyoshi, 2007; Turner et al., 1998). The impersonal context of the comp uter is thought to result in more accurate responses due to participants increased comfort answering sensitive questions. To enhance the validity of reports of peer deviancy, rather than relying on participant report of peer deviancy, future research w ould benefit from survey designs that allow participants to report the names of their closest fri ends with matching of peer and self report data (Aseltine, 1995). Peer deviancy would then be assessed by peer report of engagement in problem behaviors, reducing bias of reports. Additionally, this study evaluated a general measure of sensation seeking based on two of the four dimensions outlined by Zuckerman (2007), disinhibition and thrill/a dventure seeking. Future inves tigations of the association
131 between sensation seeking and problem behaviors would benefit from the inclusion of a more comprehensive evaluation of all four dimensi ons of sensation seeking. The construct of disinhibition would benefit in particular from consideration of closely conceptually related concepts such as lack of self control or impulsivity as well as studies of neurological functioning or biological bases of impulsive behavior. Another measurement issue is the concep tual distinction between aggression and delinquency as evaluated in the current study. The items used to assess these constructs do have some conceptual overlap which is highlighted by the observed correlation between them. Likewise, the trajectory models for both aggres sion and delinquency were very similar in terms of the number of groups identified and the shapes of the trajectori es. Despite these similarities, the distinction between these tw o behaviors is quite important and emphasizes the need to evaluate them separately. The most important distinction between aggression and delinquency is that delinquent behavior refers to illegal behavior. This includes theft, assault, and vandalism. These delinquent behaviors are more extreme and emerge later in adolescence than aggressive behaviors such as cursing, teas ing, or saying mean things to someone. As such, it remains important to evaluate aggressi on and delinquency separately to more fully understand what factors are related to transitions from less serious forms of aggressive behavior to more serious delinquent acts. In addition, altering the items used to assess aggression and delinquency to try and minimize overlap will result in scales that have not been empirically validated and which may not capture the constructs of interest. Hence, it is important to be aware of the similarity between these constructs as well as the conceptual distinctions when inte rpreting the results of the current study.
132 Finally, previous research usi ng dual trajectory analysis has on ly evaluated dual trajectory models among behaviors that occur over the same span of time (For review see Piquero, in press). The antisocial behavior s evaluated in the current st udy encompassed different time frames based on well established and validated meas ures of each behavior. Participants were asked to retrospectively report engagement in aggressive behavior over the past month while engagement in both delinquency and substance use were reported for the past year. It is unclear if the dual trajectory results presented in this stu dy would remain the same or differ if all of the behaviors were assessed over the same span of time. Implications This investigation provided some insight into the underlying mechan isms contributing to the sequential emergence of aggression, delinquenc y, and substance use during early adolescence addressing a debate within the field regardi ng the co-occurrence of these behaviors during adolescence and adulthood. Despite the fact that the co-occurrence of these antisocial behaviors is well documented in the literature, most research to date has evaluated each antisocial behavior separately or has combined them into a single construct of general deviance. However, this study took advantage of a longitudi nal research design to begin to disentangle the reciprocal influences between aggression, delinquency, a nd substance use in early adolescence. Identifying interconnections among problem behavi ors in adolescents is key to establishing effective prevention programming. The results of this study bring to light the need to take a holistic approach to preventive interventions. As shown in th is study, patterns of change in aggression, delinquency, and substance use are associated with one another during middle school. Whether these behaviors are linked to on e another causally or if they are simply different developmental manifest ations of an underlying tendency toward general deviance is still a matter of debate. It is possible that the engagement in aggressive behavior affords the
133 opportunity to develop deviant peer associations which in turn leads to a social environment conducive to continued aggressi ve behavior, escalation into delinquency, and initiation of substance use. Adolescents who are engaging in multiple antisocial behaviors represent a significant subgroup of individuals at higher risk for continued problems in adulthood. As such, it is particularly important to invest igate individual and cont extual factors that place adolescents at risk for multiple problem behaviors. This st udy evaluated both sensation seeking and deviant peer association as mechanisms underlying chan ges in aggression, delinquency, and substance use in early adolescence. This investigation shed light on the role of sensation seeking and deviant peer association on inte rconnections between antisocial behaviors rather than simply evaluating a general constr uct of deviance or indi vidual outcomes alone. By doing so, the results of this study provide a richer picture of adolescent developmen t during the middle school years which will assist in the development of more effe ctive and appropriate intervention strategies for the prevention of substance use and violence. For example, regarding peer deviancy and sensation seeking, universal programming may be useful given minimal gender and race/ethnic differences in associations with antisocial beha viors and substance use. However, it would be informative for future research to evaluate ot her adjustment outcomes (i.e., precocious sexual activity, depression) given findi ngs of unique processes among gender and race/ethnic groups found in other studies (Stanton et al., 1993). In addition, information regarding gender and racial/ethnic differences on associations between antisocial behaviors in ear ly adolescence is sparse and in consistent thus far, requiring further examination. Given evidence of gender and race/ethnic differenc es in prevalence of antisocial behaviors, as well as differential eff ects of risk and protective factors, this study
134 investigated both males and females from a vari ety of racial and ethni c backgrounds with the goal of providing the most complete pictur e of the developmental progression of, and interconnections between, antisoc ial behaviors across early ad olescence. Gaining a better understanding of commonalities and differences between genders and racial/ethnic groups in the etiology of antisocial behaviors is informa tive for the development and refinement of intervention strategies. Successful intervention techniques which promote positive youth development and minimize negative adjustment have important implications not only for individual lives but for society as a whole.
135 APPENDIX A LIFE SKILLS TRAINING HEALTH SURVEY ITEMS Substance use About how often if ever do you: 1. Smoke cigarettes 2. Drink beer, wine, wine cooler s or hard liquor (excluding us e during religious ceremonies) 3. Drink until you get drunk 4. Smoke marijuana (pot, reefer, w eed, blunts) or hashish (hash) 5. Smoke marijuana or hashish until you get high or stoned (nice) 6. Sniff glue, paint, gas or other things you inhale to get high Delinquency How many times in the past year have you: 1. Purposely damaged or destroyed property or things that di d not belong to you? 2. Thrown objects such as rocks or bottles at cars or people? 3. Picked a fight with someone? 4. Hit someone with the idea of seriously hurting them? 5. Taken something worth less than $50 that didnt belong to you? 6. Taken something from a person by for ce (other than just playing around)? 7. Beat up on someone or fought someone physical ly if they provoked you (other than just playing around)? 8. Taken something from a store wh en a clerk wasnt looking? 9. Intentionally damaged or me ssed up something in a school or some other building? 10. Taken part in a fight wher e a group of your friends were against another group? Aggression How many times in the past month have you: 1. Said mean things to someone? 2. Threatened to hurt someone? 3. Yelled at someone (you were mad at)? 4. Pushed or shoved someone on purpose? 5. Tripped someone on purpose? 6. Cursed at someone? 7. Teased someone or called someone names? 8. Argued with other people? 9. Tol d someone off? 10. Hit someone?
136 Friend Drug Use How many of your friends do you think: 1. Smoke cigarettes 2. Drink beer, wine or liquor 3. Smoke marijuana (pot, reefer, weed, blunts) 4. Use cocaine or other hard drugs 5. Sniff glue, paint, gas or other things you inhale to get high Friend Delinquency During the past year how ma ny of your friends have: 1. Ruined or damaged something on purpose that wasnt theirs. 2. Stolen something worth less than $50. 3. Hit or threatened to hit some one without any real reason. 4. Broken into some place to steal something. 5. Carried weapons. 6. Picked a fight with someone. 7. Beat someone or fought someone physically if they were provoked (other than just playing around)? Disinhibition How much do you agree or disagree with the following statements. 1. I stick to what Im doing until Im finished with it. (reverse coded) 2. I have to be reminded several times to do something 3. I am easily distracted from my work. 4. I find that I like to switch from one thing to another. 5. If I find that something is really difficult, I get frustrated and quit. 6. If I promise to do something, I can be c ounted on to deliver. (reverse coded) 7. It doesnt really take much to calm me do wn when I am excited or all wound up. (reverse coded) 8. If I ask a question, I wait for the answer rath er than jumping to the next idea. (reverse coded) 9. I have been told that I inte rrupt people in conversations. 10. In situations where I have to wait in line I can do this patie ntly. (reverse coded) 11. If I am part of a group project, I can follow suggestions of other people. (reverse coded) 12. When som eone asks me a question I usually respond with a thought ful answer. (reverse coded) 13. I often do too many things at once, in stead of concentration on one task.
137 Thrill/Adventure Seeking How much do you agree or disagree with the following statements. 1. I enjoy taking risks. 2. I would enjoy fast driving. 3. I would do almost anything on a dare. 4. I think life with no danger in it would be dull for me.
138 LIST OF REFERENCES Achenbach, T. M. (1991). Manual for the Youth Self-Report and 1991 profile Burlington, VT: University of Vermont, De partment of Psychiatry. Achenbach, T. M., McConaughy, S. H., & Howell, C. T. (1987). Child/adolescent behavioral and emotional problems: Implications of cr oss-informant correlations for situational specificity. Psychological Bulletin, 101 (2), 213-232. Aiken, L. S. & West, S. G. (1991). Multiple regression: Testi ng and interpreting interactions Newbury Park, CA: Sage Publications. Allison, P. D. (2005). Fixed effects regression methods for longitudinal data using SAS Cary, NC: SAS Institute Inc. Andrews, J. A. (2005). Substance abuse in girls. In D. J. Bell, S. L. Foster, & E. J. Mash (Eds.), Handbook of behavioral and emo tional problems in girls (pp. 181-209). New York: Kluwer Academic/Plenum Publishers. Andrews, J. A., Hampson, S. E., Barckley, M., Gerrard, M., Gibbons, F.X. (in press). The effect of early cognitions on cigarette and alcohol use in adolescence. Psychology of Addictive Behaviors. Andrews, J. A., Tildesley, E., Hops, H., Dunca n, S., & Severson, H. H. (2003). Elementary school age childrens future intentions and use of substances. Journal of Clinical Child & Adolescent Psychology, 32, 556-567. Andrucci, G. L., Archer, R. P., Pancoast, D. L ., & Gordon, R. A. (1989). The relationship of MMPI and sensation seeking scales to adolescent drug use. Journal of Personality Assessment, 53(2), 253-266. Arnett, J. (1992). Reckless behavior in a dolescence: A developmental perspective. Developmental Review, 12 (4), 339-373. Arnett, J. (1995). The young and the reck less: Adolescent reckless behavior. Current Directions in Psychological Science, 4 (3), 67-71. Ball, S. A. (1995). The validity of an alternat ive five-factor measure of personality in cocaine abusers. Psychological Assessment, 2, 148-154. Bandura, A. (1977). Social learning theory Oxford, England: Prentice-Hall. Bardo, M. T. & Mueller, C. W. (1991). Se nsation seeking and drug abuse prevention from a biological perspective. In Donohew, L ., Sypher, H. E., & Bulkoski, W. (Eds) Persuasive communication and drug abuse prevention, Hillsdale, NJ: Erlbaum, (pp. 195-207).
139 Bardo, M. T., Donohew, R. L., & Harrington, N. G. (1996). Psychobiology of novelty seeking and drug seeking behavior. Behavioral Brain Research, 77 (1-2), 23-43. Biglan, A., Hood, D., Brozovsky, P., Ochs, L., Ary, D., & Black, C. (1991). Subject attrition in prevention research. In W. Bukoski & K Leukefeld (Eds.), Drug abuse prevention research: Methodological issues NIDA Research Monograph 107 (pp. 213-223). Rockville MD: National Institute on Drug Abuse. Booth-Kewley, S., Larson, G. E., & Miyoshi, D. K. (2007). Social desirability effects on computerized and paper-and-pencil questionnaires. Computers in Human Behavior, 23(1), 463-477. Botvin, G. J. & Griffin, K. W. (2004). Life skills training: Empirical findings and future directions. Journal of Primary Prevention 25(2), 211-232. Botvin, G. J. & Scheier, L. M. (1997). Preventi ng drug abuse and violence. In D. K. Wilson, J. R. Rodrigue, & W. C. Taylor (Eds.), Health-promoting and health-compromising behaviors among minority adolescents (pp. 55-86). Washington D.C.: American Psychological Association. Botvin, G. J., Schinke, S. P., Epstein, J. A., & Diaz, T. (1994). Effectiveness of culturally focused and generic skills training appro aches to alcohol and drug abuse prevention among minority youths. Psychology of Addictive Behaviors, 8 (2), 116-127. Broidy, L. M., Nagin, D. S., Tremblay, R. E., Bates, J. E., Brame, B., Dodge, K. A., Fergussion, D., Horwood, J. L., Loeber, R., Laird, R., Lyna m, D. R., Moffitt, T. E., Pettit, G. S., & Vitaro, F. (2003). Developmental trajec tories of childhood disr uptive behaviors and adolescent delinquenc y: A six-site, cross-national study. Developmental Psychology, 39(2), 222-245. Brook, J. S., Whiteman, M., Gordon, A. S., & Cohe n, P. (1986). Dynamics of childhood and adolescent personality traits and adolescent drug use. Developmental Psychology, 22 (3), 403-414. Bronfenbrenner, U. (1979a). The ecology of human development. Cambridge: Harvard University Press. Bronfenbrenner, U. (1979b). Contexts of child rearing: Problems and prospects. American Psychologist, 34 (10), 844-850. Brown, S. A. (1993). Recovery patterns in adolescent substance abusers. In J. S. Baer, G. A. Marlatt, & R. J. McMahon (Eds.), Addictive behavior across the life-span: Prevention, treatment, and policy issues (pp. 61-163). Beverly Hills, CA: Sage. Brown, T. L., Miller, J. D., & Clayton, R. R. (2004). The generalizability of substance use predictors across racial groups. Journal of Early Adolescence, 24 (3), 274-302.
140 Catalano, R. F., & Hawkins, J. D. (1996). The so cial development model: A theory of antisocial behavior. In A. Blumstei n, & D. Farrington (Eds.), Delinquency and crime: Current theories (pp. 149 197). New York: Cambridge University Press. Centers for Disease Control, 2003. Youth Risk Behavior Surveilla nce United States, 2003. In: CDC Surveillance Summaries. U.S. Department of Health and Human Services. Centers for Disease Control, 2003. National Su rvey on Drug Use & Health United States, 2003. In: CDC Surveillance Summaries. U.S. Department of Health and Human Services. Centers for Disease Control and Prevention. 2005 Youth Risk Behavior Survey Available at: www.cdc.gov/yrbss. Accessed on April 14, 2007. Chassin, L., & Ritter, J. (2001). Vulnerabil ity to substance use disorders in childhood and adolescence. In R. E. Ingram & J. M. Price (Eds.), Vulnerability to psychopathology: Risk across the lifespan (pp. 107-134). New York: The Guilford Press. Chen, K. & Kandel, D. (1995). The natural hi story of drug use from adolescence to the midthirties in a general population sample. American Journal of Public health, 85, 41-47. Clayton, R. R., Leukefeld, C. G., Donohew, L., Bardo, M., & Harrington, N. G. (1995). Risk and protective factors: A br ief review. In: C.G. Leukefeld & R. R. Clayton (Eds), Prevention Practice in Substance Abuse The Haworth Press, (pp. 7-17). Cohen, J., & Cohen, P. (1983). Applied multiple regression/ correlation analysis for the behavioral sciences: Second Edition Hillsdale, NJ: Lawrence Erlbaum. Conduct Problems Prevention Research Group. (1992) A developmental and clinical model for the prevention of conduct disord ers: The Fast Track Program. Development and Psychopathology, 4, 509-527. Conduct Problems Prevention Research Group. (2004). The effects of the Fast Track program on serious problem outcomes at th e end of elementary school. Journal of Clinical Child and Adolescent Psychology, 33, 650-661. Crick, N. R., & Dodge, K. A. (1994). A review and reformulation of social information-processing mechanisms in children's social adjustment. Psychological Bulletin, 115 74-101. Crick, N. R. & Grotpeter, J. K. (1995). Rela tional aggression, gender, and social-psychological adjustment. Child Development, 66 (3), 710-722.
141 Curran, P. J. & Muthen, B. O. (1999). The app lication of latent curv e analysis to testing developmental theories in intervention research. American Journal of Community Psychology, 27(4), 567-595. Day, N. L., Wagener, D. K., & Taylor, P. M. (1985). Measurement of substance use during pregnancy: Methodologic issues. NIDA Research Monograph, 59 36. Dishion, T. J. & Loeber, R. (1985). Adolescent ma rijuana and alcohol use: The role of parents and peers revisited. American Journal of Dr ug and Alcohol Abuse, 11 11-25. Dishion, T. J. & Patterson, G. R. (2006). The development and ecology of antisocial behavior in children and adolescents. In D. Cicchetti & D. J. Cohen (Eds.), Developmental Psychopathology: Risk, disorder, and adaptation (2nd ed., pp. 503-541). New Jersey: John Wiley & Sons. Dodge, K. A., Coie, J. D., & Lynam, D. (2006). Aggression and antisocial behavior in youth. In N. Eisenberg, W. Damon, & R. M. Lerner (Eds.), Handbook of child psychology: Social, emotional, and personality development (6th ed., pp. 719-788). New Jersey: John Wiley & Sons. Donohew, R. L., Hoyle, R. H., Clayton, R. R., Ski nner, W., Colon, S. E., & Rice, R. E. (1999). Sensation seeking and drug use by adolescents and their friends: Models for marijuana and alcohol. Journal of Studies on Alcohol, 60, 622-631. Donohew, R. L., Lorch, E., & Palmgreen, P. (1 991). Sensation seek ing and targeting of televised anti-drug PSAs. In Donohew, L., Sypher, H. E., & Bulkoski, W. (Eds) Persuasive communication and drug abuse prevention, Hillsdale, NJ: Erlbaum, (pp. 209226). Eide, A. H., Acuda, S. W., Khan, N., Aaroe, L. E., & Loeb, M. E. (1997). Combining cultural, social and personality trait variables as predic tors for drug use among adolescents in Zimbabwe. Journal of Adolescence, 20 (5), 511-524. Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use Beverly Hills, CA: Sage Publications. Elliott, D., Huizinga, D., & Menard, S. (1989). Multiple problem youth: Delinquency, substance use, and mental health problems. New York: Springer-Verlag. Ensminger, M. E. & Juon, H. S. (1998). Tran sition to adulthood among high-risk youth. In R. Jessor (Ed.), New perspectives on adol escent risk behavior (pp. 365-391). New York: Cambridge University Press. Evans, R. I., Hansen, W. B., & Mittlemark, M. B. (1977). Increasing the validity of self-reports of behavior in a smoking in children investigation. Journal of Applied Psychology, 62 (4), 521-523.
142 Eysenck, H. J. & Eysenck, S. B. G. (1975). Manual of the Eysenck personality questionnaire. London: Hodder & Stoughton. Farrell, A. D., Sullivan, T. N., Esposito, L. E., Meye r, A. L., & Valois, R. F. (2005). A latent growth curve analysis of th e structure of aggression, drug use, and delinquent behaviors and their interrelations over time in urban and rural adolescents. Journal of Research on Adolescence, 15 (2), 179-204. Fischman, M. W., Johanson, C. E. (1998). Ethical and practical issues involved in behavioral pharmacology research that administers drugs of abuse to human volunteers. Behavioural Pharmacology, 9 (7), 479. Friedman, A. S. & Glassman, K. (2000). Family risk factors versus peer risk factors for drug abuse: A longitudinal study of an Afri can American urban community sample. Journal of Stustance Abuse Treatment, 18 267-275. Gottfredson, M. R. & Hirschi, T. (1990). A general theory of crime Stanford, CA: Stanford University Press. Graber, J. A. (1999). Sequence and timing among young adult transitions: The impact of poverty on developmental course. In A. Booth, A. C. Crouter & M. J. Shanahan (Eds.), Transitions to adulthood in a changing econom y: No work, no family, no future? (pp. 234-243). Westport, CT: Praeger. Graber, J. A. & Brooks-Gunn, J. (1996). Transiti ons and turning points: Navigating the passage from childhood through adolescence. Developmental Psychology, 32 (4), 768-776. Graber, J. A. & Brooks-Gunn, J. (1999). Developmental transitions: Linking human development with tobacco prevention research. Nicotine & Tobacco Research, 1 S73S77. Graber, J. A., Brooks-Gunn, J., & Petersen, A. C. (Eds.). (1996). Transitions through adolescence: Interperso nal domains and context Mahwah, NJ: Lawrence Erlbaum & Associates. Graber, J. A., Lynne, S. D., Nichols, T. R., Brooks-Gunn, J., & Botvin, G. J. (2007). Testing mechanisms for pubertal timing effects on drug use in urban, minority girls and boys Manuscript in preparation. Graber, J. A., Nichols, T. R., Lynne, S. D ., Brooks-Gunn, J., & Botvin, G. J. (2006). A longitudinal examination of family, friend, and media influences on competent versus risky behaviors among urban minority youth. Applied Developmental Science, 10 7585.
143 Hampson, S. E., Andrews, J. A., Barckley, M., & Severson, H. H. (2006). Personality predictors of the development of elementary-school childrens intent ions to drink alcohol: The mediating effects of at titudes and subjective norms. Psychology of Addictive Behaviors, 20(3), 288-297 Hansen, W.B., Collins, L.M., Malotte, C.K., J ohnson, C.A., & Fielding, J.E. (1985). Attrition in prevention research. Journal of Behavi oral Medicine, 8 261 275. Hartup, W. W. (1974). Aggression in childhood: Developmental perspectives. American Psychologist, 29 (5), 336-341. Haviland, A., Nagin, D. S., & Rosenbaum, P. R. (2007). Combining propensity score matching and group-based trajectory analys is in an observational study. Psychological Methods, 12(3), 247-267. Huba, G. J., Newcomb, M. D., & Bentler, P. M. (1981). Comparison of canonical correlation and interbattery factor analysis on se nsation seeking and drug use domains. Applied Psychological Measurement, 5, 291-306. Jessor, R. (1987). Problem-behavior theory, psychosocial development, and adolescent problem drinking. British Journal of Addiction, 82 331-342. Jessor, R. (1992). Risk behavi or in adolescence: A psychos ocial framework for understanding and action. Developmental Review, 12 (4), 374-390. Johnston, L. D., OMalley, P. M., Bachman, J. G., & Schulenberg, J. E. (2006). Monitoring the Future national results on adolescent drug use: Overview of key findings, 2005. (NIH Publication No. 06-5882). Bethesda, MD: National Institute on Drug Abuse. Junger-Tas, J., Marshall, I. H., Ribeaud, D. (2003). Delinquency in an international perspective: The international se lf-reported delinquency study (ISRD) Monsey, NY: Criminal Justice Press. Kandel, D. B. & Andrews, K. (1987). Processes of adolescent socialization by parents and peers. International Journal of the Addiction, 22, 319-342. Kandel, D. B., Kessler, R. C., & Margulies, R. Z. (1978). Antecedents of adolescent initiation into stages of drug use: A developmental analysis. Journal of Youth and Adolescence, 7(1), 13-40. Kandel, D. B. & Logan, J. (1984). Patterns of drug use from adolescence to young adulthood: Periods of risk for initiation, c ontinued use, and discontinuation. American Journal of Public Health, 74 660-666.
144 Kass, R. E. & Wasserman, L. (1995). A reference Bayesian test for nested hypotheses and its relationship to the Schwarz criterion. Journal of the American St atistical Association, 90 928-934. Laventure, M., Dry, M., & Pau z, R. (2006). Consommation de psychotropes et persistence du trouble des conduits chez les adolescents. Alcoologie et Addictologie 28(3), 223-230. Lillehoj, C. J., Trudeau, L., Spoth, R., & Madon, S. (2005). Externalizing behaviors as predictors of substance initiation trajectories among rural adolescents. Journal of Adolescent Health, 37 493-501. Loeber, R. (1988). Natural histories of conduct problems, delinquency, and associated substance use: Evidence for developmental progr essions. In B. B. Lahey & A. E. Kazdin (Eds.), Advances in clinical child psychology (pp. 73-124). New York: Plenum Press. Loeber, R. & Hay, D. (1997). Key issues in the development of aggression and violence from childhood to early adulthood. Annual Review of Psychology, 48 371-410. Lynam, D. R., Milich, R., Zimmerman, R., Novak, S. P., Logan, T. K., Martin, C., et al. (1999). Project DARE: No effect s at 10-year follow-up. Journal of Consulting and Clinical Psychology, 67(4), 590-593. Lynne, S. D., Graber, J. A., Nichols, T. R., Brooks-Gunn, J., & Botvin, G. J. (2007). Links between pubertal timing, peer influences and externalizing behaviors among urban students followed through middle school Journal of Adolescent Health, 40, 181.e7181.e13. Margulies, R. Z., Kessler, R. C., & Kandel, D. B. (1977). A longitudinal study of onset of drinking among high-school students. Journal of studies on alcohol, 38 (5), 897-912. Mason, W. A. & Windle, M. (2002 ). Reciprocal relations betw een adolescent substance use and delinquency: A longitudinal la tent variable analysis. Journal of Abnormal Psychology, 111(1), 63-76. Mayes, L. C. & Suchman, N. E. (2006). Develo pmental pathways to substance abuse. In D. Cicchetti & D. J. Cohen (Eds.), Developmental Psychopathology : Risk, disorder, and adaptation (2nd ed., pp. 599-619). New Jersey: John Wiley & Sons. McCord, J. (1990). Problem behaviors. In S. S. Feldman & G. R. Elliott (Eds.), At the threshold: The developing adolescent (pp. 414-430). Cambridge, MA: Harvard University Press. Moffitt, T. E. (2006). Life-course-persistent vers us adolescence-limited antisocial behavior. In D. Cicchetti & D. J. Cohen (Eds.), Developmental Psychopathology: Risk, disorder, and adaptation (2nd ed., pp. 570-598). New Jersey: John Wiley & Sons.
145 Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. (2001). Sex differences in antisocial behaviour: Conduct disorder, delinquency, and violence in the Dunedin Longitudinal Study. United Kingdom: Cambridge University Press. Nagin, D. S. (1999). Analyzing developmental trajectories: A semiparametric, group-based approach. Psychological Methods, 4 (2), 139-157. Nagin, D. S. (2005). Group-based modeling of development Cambridge, MA: Harvard University Press. Nagin, D. S. & Tremblay, R. E. (2001a). Anal yzing developmental traject ories of distinct but related behaviors: A group-based method. Psychological Methods, 6 (1), 18-34. Nagin, D. S. & Tremblay, R. E. (2001b). Parent al and early childhood pred ictors of persistent physical aggression in boys from kindergarten to high school. Archives of General Psychiatry, 58 (4), 678-684. Nagin, D. S. & Tremblay, R. E. (2005a). Developmental trajectory groups: Fact or a useful statistical fiction? Criminology, 43 (4), 873-904. Nagin, D. S. & Tremblay, R. E. (2005b). From seduction to passion: A response to Sampson and Laub. Criminology, 43 (4), 915-918. Newcomb, M. D. (1995). Drug use etiology among ethnic minority adolescents: Risk and protective factors. In G. J. Botvin, S. P. Schinke, & M. A. Orlandi (Eds.), Drug abuse prevention with multiethnic youth (pp. 105-129). Thousand Oaks, CA: Sage. Newcomb, M. D. (1997). Psychosocial pred ictors and consequences of drug use: A developmental perspective within a prospective study. Journal of Addictive Diseases, 16(1), 51-89. Newcomb, M. D. & Bentler, P. M. (1989). Substance use and abuse among children and teenagers. American Psychologist, 44 (2), 242-248. Newcomb, M. D. & Earleywine, M. (1996). Intraper sonal contributors to drug use: The willing host. American Behavioral Scientist, 39 (7), 823-837. Newcomb, M. D. & McGee, L. (1989). Adolescen t alcohol use and other delinquent behaviors: A one year longitudinal analysis co ntrolling for sensation seeking. Criminal Justice and Behavior, 16, 345-369. Newcomb, M. D. & McGee, L. (1991). Influen ce of sensation seeking on general deviance and specific problem behaviors from adolescence to young adulthood. Journal of Personality and Social Psychology, 61 (4), 614-628.
146 Nichols, T. R., Graber, J. A., Brooks-Gunn, J., & Botvin, G. J. (2006). Sex differences in overt aggression and delinquency among urba n minority middle school students. Journal of Applied Developmental Psychology, 27 (1), 78-91. Odgers, C.L. & Moretti, M.M. (2002). Aggressive and antisocial girls: Research update and challenges, International Journal of Fo rensic Mental Health, 1 (2), 103-119. Patterson, G. R., DeBaryshe, B. D., & Ramsey, E. (1989). A developmental perspective on antisocial behavior. American Psychologist, 44 (2), 329-335. Phares, V., Compas, B. E., & Howell, D. C. ( 1989). Perspectives on chil d behavior problems: Comparisons of childrens self-reports with parent and te acher reports. Psychological Assessment, 1(1), 68-71. Piquero, A. R. (In press). Taking stock of deve lopmental trajectories of criminal activity over the life course. In A. Liberman (Ed.), The long view of crime (pp. 23-78). Springer. Piquero, A. R., Farrington, D. P., & Blumstein, A. (2007). Key issues in criminal career research: New analyses of the Camb ridge Study in Delinquent Development New York: Cambridge University Press. Rosenbaum, M. (1980). A schedule for assessing self-control behaviors: Preliminary findings. Behavior Therapy, 11 (1), 109-121. Sampson, R. J. & Laub, J. H. (2005). Seduc tions of method: Rejoinder to Nagin and Tremblays Developmental trajecto ry groups: Fact or fiction? Criminology, 43(4), 905-913. Scheier, L. M. & Botvin, G. J. (1996). Purpos e in life, cognitive efficacy, and general deviance as determinants of drug abuse in urban black youth. Journal of Child & Adolescent Substance Abuse, 5 (1), 1-26. Sher, K. J. (1991). Children of alcoholics: A critica l appraisal of theory and research. Chicago: University of Chicago Press. Simcha-Fagan, O., Gersten, J. C., & Langner, T. S. (1986). Early precursors and concurrent correlates of patterns of illic it drug use in adolescence. The Journal of Drug Issues, 16(1), 7-28. Snow, W. H., Gilchrist, L. D., & Schinke, S. P. (1985). A critique of progress in adolescent smoking prevention. Child & Youth Services, 7 (1), 1-19. Stacy, A. W., & Newcomb, M. (1995). Long-term social-psychological influences on deviant attitudes and criminal behavi or. In H. B. Kaplan (Ed.), Drugs, crimes, and other deviant adaptations: Longitudinal studies (pp. 99-128). New York: Plenum Press.
147 Stanton, B., Romer, D., Ricardo, I., Black, M., Feigelman, S., & Galbraith, J. (1993). Early initiation of sex and its lack of associati on with risk behaviors among adolescent AfricanAmericans. Pediatrics, 92 (1), 13-19. Svensson, R. (2003). Gender differences in ad olescent drug use: The impact of parental monitoring and peer deviance. Youth & Society, 34 (3), 300-329. Tabachnick, B. G. & Fidell, L. S. (2007). Using Multivariate statistics (5th ed.). New York: Pearson Education. Tiet, Q. Q., Wasserman, G. A., Loeber, R., McReynolds, L. S., & Miller, L. S. (2001). Developmental and sex differences in types of conduct problems. Journal of Child and Family Studies, 10(2), 181-197. Tourangeau, R., Smith, T. W., & Rasinski, K. A. (1997). Motivation to report sensitive behaviors on surveys: Evidence from a bogus pipeline experiment. Journal of Applied Social Psychology, 27 (3), 209-222. Turner, C. F., Ku, L., Rogers, S. M., Lindberg, L. D., & Pleck, J. H. (1998). Adolescent sexual behavior, drug use, and violence: Increased re porting with computer survey technology. Science, 280 (5365), 867-873. Vitaro, F., Brendgen, M., Ladouceur, R., & Trem blay, R. E. (2001). Gambling, delinquency, and drug use during adolescence: Mutual influences and common risk factors. Journal of Gambling Studies, 17(3), 171-190. White, H. R., Labouvie, E. W., & Bates, M. E. (1985). The relationship between sensation seeking and delinquency: A longitudinal analysis. Journal of Research in Crime and Delinquency, 22, 197-211. Zuckerman, M. (1979). Sensation seeking: Beyond th e optimal level of arousal. Hillsdale, NJ: Erlbaum. Zuckerman, M. (1986). Sensation seeking and the endogenous deficit theory of drug abuse National Institute on Drug Abuse Research Monograph Series No. 74. Neurobiology of Behavioral Control in Drug Abuse. Depart ment of Health and Human Services, pp. 5970. Zuckerman, M. (2007). Sensation seeking and risky behavior Washington D.C.: American Psychological Association. Zuckerman, M., Ball, S., & Black, J. (1990). In fluences of sensation seeking, gender, risk appraisal, and situationa l motivation on smoking. Addictive Behaviors, 15 (3), 209-220.
148 BIOGRAPHICAL SKETCH Sarah Delphia Lynne was born on September 8, 1979, in Fairfax, Virginia, to Karen Barrett-Perry. She has two younger siblings, her brother Jonathan and her sister Elyse. While her father was never a parental figur e in her life, Sarahs mother wa s a diligent and driven woman who always provided her children with not only th e basic material needs but also a very warm and supportive home. Sarah learned the values of independence, patience, and persistence during childhood and adolescence by watching her mo ther successfully overcome adversity. She also learned the value and im portance of family. She was married to Matthew David Russ Landsman on October 20th, 2007. Her husband is an honest and supportive man whose willingness to compromise and pervasive good natu re have contributed tremendously to Sarahs happiness and success. After graduating high school, Sarah went on to Lord Fair fax Community College where she earned an Associate of Arts and Science degree. She then transferred to Virginia Polytechnic Institute and State University where she earned a Bachelor of Science in Psychology. While at Virginia Tech, Sarah worked with elementary age students enhancing lite racy and math skills, which contributed, in part, to her desire to study human development. She applied and was accepted into the developmental psychology graduate program at the University of Florida, where she is currently developing her skills as an academician and researcher. Her research interests include examining the psychological, bi ological, and social f actors that impact the development of healthy/adaptive behaviors in adolescence and minimize maladaptive/pathological outcomes. She receiv ed her Master of Science degree in May 2005 and her Ph.D. in Developmental Psychology in May 2008.